Cleaning mode switching method and device for multifunctional swimming pool cleaning robot

Abstract

The invention relates to the technical field of swimming pool cleaning equipment control, and provides a cleaning mode switching method and device for a multifunctional swimming pool cleaning robot. According to the method, water quality data corresponding to a swimming pool, stain distribution images of pool walls and pool bottoms, swimming pool structure size data and stain type data are collected in real time; dividing a cleaning area according to the swimming pool structure size data, generating a stain area marking result in the cleaning area in combination with the stain distribution image and the stain type data, and judging the pollution level of the whole swimming pool according to the water quality data so as to retrieve a corresponding cleaning mode from a preset cleaning mode library, the preset cleaning mode library comprises a global coverage mode, a key enhancement mode, a corner fitting mode and a water quality maintenance mode; and according to the called cleaning mode, a corresponding cleaning path plan and execution instruction is generated, and the driving module, the scrubbing module and the purification module are controlled to work cooperatively. According to the method, accurate analysis of the swimming pool environment is achieved, and blind cleaning of a traditional fixed mode is avoided.

Description

Technical Field

[0001] This application relates to the field of pool cleaning equipment control technology, and in particular to a method and device for switching cleaning modes of a multi-functional pool cleaning robot. Background Technology

[0002] Pool cleaning robots are commonly used equipment for pool maintenance, and their cleaning efficiency and adaptability directly affect the effectiveness of pool maintenance. Traditional pool cleaning robots typically have preset fixed cleaning modes (such as fixed path coverage, single-frequency brushing, etc.), but in actual pool environments, water quality, stain distribution (such as scale on pool walls, sediment on pool bottom, algae in corners, etc.), and pool structure (such as ladder gaps, curved pool walls) vary significantly, and the stain removal effect changes dynamically during the cleaning process. Current technologies lack the comprehensive perception capabilities of real-time water quality data, stain types, regional contamination levels, and pool structure, and cannot dynamically adjust cleaning strategies according to the real-time environment. For example, traditional solutions struggle to distinguish between lightly and heavily soiled areas and match differentiated cleaning intensities. Cleaning complex structural areas (such as the junction of pool walls and bottom) often results in insufficient contact due to the fixed robot angle, and the robots cannot intelligently switch modes based on real-time feedback data during the cleaning process, easily leading to problems of "over-cleaning" or "under-cleaning." Furthermore, existing technologies do not integrate water quality maintenance (such as the release of purifiers and circulating filtration) with physical scrubbing to form a coordinated strategy, resulting in a single cleaning function.

[0003] Therefore, there is an urgent need for a method that can sense the environment in real time, intelligently divide cleaning areas, and dynamically switch cleaning modes to improve the adaptability and cleaning efficiency of pool cleaning robots. Summary of the Invention

[0004] This application provides a cleaning mode switching method and device for a multi-functional swimming pool cleaning robot, which aims to solve the problem that traditional solutions have difficulty distinguishing between lightly soiled areas and heavily soiled areas and matching different cleaning intensities. When cleaning complex structural areas (such as the junction of the pool wall and the pool bottom), the fixed angle of the robot body often leads to insufficient contact, and it is impossible to intelligently switch modes based on real-time feedback data during the cleaning process, which easily leads to problems of "over-cleaning" or "under-cleaning".

[0005] In a first aspect, this application provides a cleaning mode switching method for a multi-functional pool cleaning robot, applied to the cleaning robot, including: The system uses a water quality sensor, camera, ultrasonic ranging module, and stain recognition module to collect real-time data on the pool's water quality, stain distribution on the pool walls and bottom, pool structure dimensions, and stain types. The cleaning area is divided based on the pool's structural dimensions. Combined with stain distribution images and stain type data, stain area markings are generated within each cleaning area. The overall pollution level of the pool is determined based on water quality data, and a corresponding cleaning mode is retrieved from a pre-set cleaning mode library. This library includes a global coverage mode, a focused cleaning mode, a corner-fitting mode, and a water quality maintenance mode. The global coverage mode uses a spiral path for uniform cleaning. The focused cleaning mode uses a repetitive, high-frequency brushing method to clean heavily soiled areas. The corner-fitting mode adjusts the machine angle and uses a flexible brush head for cleaning the junction of the pool wall and bottom, and the gaps in the ladder. The water quality maintenance mode releases a water purifier and performs circulating filtration when water quality data is abnormal. Based on the retrieved cleaning mode, the corresponding cleaning path plan and execution instructions are generated to control the drive module, scrubbing module and purification module to work together. During the cleaning process, real-time status data of the cleaning area is acquired, and the cleaning area division and cleaning mode parameters are dynamically adjusted based on the real-time status data. If the current cleaning mode is found to be not performing up to the preset cleaning standard, a matching cleaning mode is retrieved from the cleaning mode library.

[0006] In some embodiments, the real-time acquisition of water quality data, stain distribution images of the pool walls and bottom, pool structure size data, and stain type data via the equipped water quality sensor, camera, ultrasonic ranging module, and stain recognition module includes: controlling the water quality sensor to continuously acquire data on turbidity, pH value, residual chlorine concentration, and pollutant particle content of the pool water; controlling the camera to acquire images of the pool walls and bottom at a preset frame rate, generating continuous frame images containing the location and shape of stains; controlling the ultrasonic ranging module to emit ultrasonic signals to structures such as the pool walls, bottom, and ladders, calculating the distance data of each structure based on the echo time, and constructing a three-dimensional outline of the pool; and controlling the stain recognition module to identify the stain type as any one of mud, algae, scale, and grease based on image color features, texture features, and grayscale value differences, and marking the coordinate position of each stain type in the image.

[0007] In some embodiments, before dividing the cleaning area according to the pool structure size data, the method further includes: performing noise reduction processing on the water quality data using a median filtering algorithm to remove sudden outliers; using a histogram equalization algorithm to enhance the contrast of the stain distribution image, improving the distinction between stains and the pool wall or bottom background; performing coordinate calibration on the pool structure size data obtained by the ultrasonic ranging module, establishing a local coordinate system with the initial position of the cleaning robot as the coordinate origin, and correcting measurement errors through multi-frame data fusion; and forming a standardized vector based on the preprocessed water quality data, stain distribution image, pool structure size data, and stain type data.

[0008] In some embodiments, the step of dividing the cleaning area according to the pool structure size data includes: dividing the pool into a bottom plane area, a vertical pool wall area, a circular arc area where the pool wall and bottom meet, a ladder area, and a handrail gap area based on the calibrated three-dimensional contour data of the pool; assigning a unique area identifier to each area, and setting the upper limit of the moving speed of the cleaning robot and the allowable range of the body tilt angle in the area according to the geometric characteristics of the area; for irregular pools, decomposing the complex structure into several calculable standard geometric units through a piecewise fitting algorithm, and dividing them into cleaning sub-areas respectively; the irregular pool includes curved pool walls and irregular steps.

[0009] In some embodiments, generating stain area marking results within the clean area by combining stain distribution images and stain type data includes: extracting stain contours from the preprocessed stain distribution image using an image segmentation algorithm, and calculating the area and average pixel grayscale value of each stain area; calculating a stain index for each stain area based on a preset stain degree weight for the stain type, combined with the stain area and average grayscale value; comparing the stain index with preset threshold ranges for heavy stain areas, moderate stain areas, and light stain areas, marking heavy stain areas, moderate stain areas, and light stain areas within the corresponding clean area, and attaching a stain type label to each marked area; wherein, the stain index corresponding to a heavy stain area is greater than a first threshold, the stain index corresponding to a moderate stain area is less than or equal to the first threshold and greater than a second threshold, the stain index corresponding to a light stain area is less than or equal to the second threshold, and the first threshold is greater than the second threshold.

[0010] In some embodiments, the preset cleaning standards include a residual stain area ratio of no more than 5% for lightly soiled areas, no more than 10% for moderately soiled areas, and no more than 20% for heavily soiled areas, and a decrease in key indicators in the water quality data of no less than 30%, including turbidity and pollutant particle content; if the current cleaning mode is found to be not meeting the preset cleaning standards, a matching cleaning mode is retrieved from the cleaning mode library, including: after the cleaning mode is completed or after cleaning each sub-area, comparing the real-time status data with the preset cleaning standards, if the residual area ratio or water quality indicator improvement of any type of stain area does not meet the standard, triggering the mode switching logic; wherein, the mode switching logic includes recalculating the cleaning priority and retrieving the matching cleaning mode based on the currently collected real-time water quality data, residual stain distribution image and pool structure data, and if the global coverage mode does not meet the standard, the key enhancement mode is retrieved to treat the residual area.

[0011] In some embodiments, determining the overall pollution level of the swimming pool based on water quality data to retrieve the corresponding cleaning mode from a preset cleaning mode library includes: establishing a water pollution level judgment rule library, which defines multi-level threshold ranges for turbidity, pH value, residual chlorine concentration, and pollutant particle content; using a fuzzy logic algorithm to fuzzify the pollution degree of each water quality parameter to generate fuzzy membership values; weighting and summing the fuzzy membership values ​​of each parameter using preset weighting coefficients to obtain the overall pollution level of the swimming pool; when the overall pollution level reaches a preset activation threshold, prioritizing the retrieval of the water quality maintenance mode and adjusting the purifying agent release amount and circulation filtration time according to the pollution level.

[0012] In some embodiments, generating corresponding cleaning path planning and execution instructions based on the retrieved cleaning mode, and controlling the drive module, brushing module, and purification module to work collaboratively, includes: if the global coverage mode is retrieved, generating a spiral coverage path starting from the current position of the cleaning robot based on the coordinate range of the bottom plane area, with the path spacing not exceeding the effective cleaning width of the brushing module, and controlling the drive module to move at a constant speed; if the focused reinforcement mode is retrieved, for the marked heavily soiled areas, generating a reciprocating cleaning sub-path by extending 10 centimeters outward from the area boundary coordinates, and controlling the brush head of the brushing module. The rotation speed is increased to a high-frequency setting, increasing the contact pressure between the brush head and the dirt surface. If the corner fitting mode is selected, the tilt angle of the cleaning robot body is calculated based on the radius of curvature of the arc area where the pool wall and bottom meet, so that the brush head axis is perpendicular to the tangent of the arc. The flexible brush head is then switched, and the drive module is controlled to move along the junction line at a low speed. If the water quality maintenance mode is selected, the corresponding type of purifier is released from the purifier storage compartment according to the type of pollutant in the water quality data. The built-in filter pump is activated to circulate and filter the water at a flow rate 1.5 times that of the pool water circulation flow rate, and the uniformity of the purifier diffusion is monitored in real time by a camera.

[0013] In some embodiments, the step of dynamically adjusting the cleaning area division and cleaning mode parameters based on real-time status data includes: acquiring images of the cleaned area in real time using a camera, calculating the difference between the images and the images of the stain distribution before cleaning, and identifying the location of any remaining stains; if the area of ​​the remaining stains accounts for more than 15% of the area of ​​the original marked area, remarking the remaining area corresponding to the location of the remaining stains as a new medium-stain area, and adding a corresponding sub-area to the existing cleaning area division; and dynamically adjusting the moving speed and brush head pressure of the cleaning robot in complex structural areas based on the real-time transmitted current value of the drive module and the torque value of the brush head motor, wherein when the current value exceeds 120% of the rated current, the moving speed is automatically reduced by 50% and the brush head pressure is reduced by 30%.

[0014] Secondly, this application provides a cleaning mode switching device for a multi-functional pool cleaning robot, applied to the cleaning robot, comprising: The data acquisition unit is used to collect real-time water quality data, images of stain distribution on the pool walls and bottom, pool structural dimensions, and stain type data of the swimming pool through the built-in water quality sensor, camera, ultrasonic ranging module, and stain recognition module. The pattern acquisition unit is used to divide the cleaning area according to the pool's structural dimensions, combine the stain distribution image and stain type data, generate stain area marking results within the cleaning area, and determine the overall pollution level of the pool based on water quality data. It then retrieves the corresponding cleaning mode from a preset cleaning mode library, which includes a global coverage mode, a focused cleaning mode, a corner-fitting mode, and a water quality maintenance mode. The global coverage mode is used for uniform coverage cleaning along a spiral path; the focused cleaning mode is used for cleaning heavily soiled areas with repetitive high-frequency brushing; the corner-fitting mode is used to adjust the machine angle and activate a flexible brush head for cleaning the junction of the pool wall and bottom and the gaps in the ladder; and the water quality maintenance mode is used to release a water purifier and perform circulating filtration when water quality data is abnormal. The instruction execution unit is used to generate corresponding cleaning path planning and execution instructions based on the retrieved cleaning mode, and control the drive module, brushing module and purification module to work together. The mode switching unit is used to acquire real-time status data of the cleaning area during the cleaning process, dynamically adjust the cleaning area division and cleaning mode parameters based on the real-time status data, and if it is detected that the current cleaning mode does not meet the preset cleaning standard, it retrieves a matching cleaning mode from the cleaning mode library.

[0015] In existing technologies, the switching of cleaning modes by pool cleaning robots largely relies on preset programs or simple sensor feedback (such as single distance detection or stain recognition). They lack multi-dimensional data fusion processing of water quality data, stain distribution images, pool structure dimensions, and stain types, and further fail to construct an intelligent algorithm model based on fuzzy logic and decision tree algorithms to achieve cleaning priority calculation and dynamic mode matching. This invention, through its technical solutions in steps one through six, proposes for the first time to utilize multiple sensors to collect multi-dimensional data in real time, dynamically divide cleaning areas, mark stain levels, and match differentiated cleaning modes through an intelligent algorithm model. Simultaneously, it dynamically adjusts strategies based on real-time feedback data during the cleaning process, solving the core problems of fixed modes and poor adaptability in traditional solutions. Existing technologies have not disclosed a similar dynamic switching mechanism combining multi-source data fusion and intelligent algorithms, nor have they proposed a control method that links global coverage, focused reinforcement, corner fitting, and water quality maintenance modes with real-time environmental data.

[0016] This application provides a cleaning mode switching method and device for a multi-functional swimming pool cleaning robot. The method collects data in real time through a water quality sensor, camera, ultrasonic ranging module, and stain recognition module. Combined with an intelligent model constructed using fuzzy logic and decision tree algorithms, it achieves accurate analysis of the swimming pool environment, avoiding the blind cleaning of traditional fixed modes. By dividing the pool into areas according to the stain level (heavy, medium, light) and pool structure, it selectively calls global coverage, key reinforcement, and corner fitting modes, improving cleaning efficiency while avoiding resource waste. By transmitting data back in real time during the cleaning process and dynamically adjusting cleaning parameters, it automatically triggers mode switching when the effect is not up to standard, ensuring the reliability of the cleaning effect. By combining physical scrubbing (different brush heads, paths) with water quality maintenance (purifying agent release, circulating filtration), a multi-functional collaborative cleaning system is formed, covering all scenarios of swimming pool maintenance needs.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic flowchart illustrating the steps of a cleaning mode switching method for a multi-functional pool cleaning robot according to an embodiment of this application; Figure 2 This is a schematic diagram illustrating the principle of a cleaning mode switching method for a multi-functional swimming pool cleaning robot provided in one embodiment of this application.

[0020] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0023] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0024] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0025] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0026] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0027] Pool cleaning robots are commonly used equipment for pool maintenance, and their cleaning efficiency and adaptability directly affect the effectiveness of pool maintenance. Traditional pool cleaning robots typically have preset fixed cleaning modes (such as fixed path coverage, single-frequency brushing, etc.), but in actual pool environments, water quality, stain distribution (such as scale on pool walls, sediment on pool bottom, algae in corners, etc.), and pool structure (such as ladder gaps, curved pool walls) vary significantly, and the stain removal effect changes dynamically during the cleaning process. Current technologies lack the comprehensive perception capabilities of real-time water quality data, stain types, regional contamination levels, and pool structure, and cannot dynamically adjust cleaning strategies according to the real-time environment. For example, traditional solutions struggle to distinguish between lightly and heavily soiled areas and match differentiated cleaning intensities. Cleaning complex structural areas (such as the junction of pool walls and bottom) often results in insufficient contact due to the fixed robot angle, and the robots cannot intelligently switch modes based on real-time feedback data during the cleaning process, easily leading to problems of "over-cleaning" or "under-cleaning." Furthermore, existing technologies do not integrate water quality maintenance (such as the release of purifiers and circulating filtration) with physical scrubbing to form a coordinated strategy, resulting in a single cleaning function.

[0028] Therefore, there is an urgent need for a method that can sense the environment in real time, intelligently divide cleaning areas, and dynamically switch cleaning modes to improve the adaptability and cleaning efficiency of pool cleaning robots.

[0029] Please see Figure 1 , Figure 1 This is a schematic flowchart illustrating a cleaning mode switching method for a multi-functional swimming pool cleaning robot according to an embodiment of this application. This cleaning mode switching method can be implemented by the cleaning robot itself, and this application embodiment does not limit the specific type of cleaning robot provided.

[0030] Specifically, such as Figures 1 to 2 As shown, the cleaning mode switching method of the provided multi-functional pool cleaning robot includes steps S101 to S104, which are detailed below: Step S101. Using the built-in water quality sensor, camera, ultrasonic ranging module and stain recognition module, the water quality data, stain distribution images of the pool wall and bottom, pool structure size data and stain type data are collected in real time.

[0031] Specifically, by integrating multiple sensors, multi-dimensional data on the pool environment is acquired in real time, providing a basis for subsequent strategy formulation. The collected data includes: water quality data: degree of water pollution, pH value, turbidity, residual chlorine content, etc.; stain distribution and type: location, area, severity (mild / severe), and type (scale, silt, algae, etc.) of stains on the pool walls / bottom; pool structural dimensions: geometric parameters of pool wall curvature, ladder location, pool bottom slope, and the junction area (corners) between the pool walls and bottom; and dynamic feedback data: real-time updates on stain removal effectiveness and water quality changes during the cleaning process.

[0032] Water quality sensors employ multi-parameter water quality sensors (such as pH electrodes, turbidity sensors, and conductivity sensors) and are deployed at the front or side of the robot to monitor water quality parameters in the water flow in real time.

[0033] The camera is equipped with a waterproof high-definition camera (supporting infrared or underwater supplemental lighting), which is installed on the front or top of the device to capture images of the pool walls, bottom, and corners. The resolution is ≥1080P, the frame rate is ≥30fps, and it supports a wide-angle lens to cover a large area.

[0034] The ultrasonic ranging module deploys 3-6 ultrasonic sensors around the body to emit fan-shaped sound waves to scan the pool walls, ladders, corners and other structures, acquire distance data (accuracy ±5mm), and construct a three-dimensional outline model of the pool.

[0035] The stain recognition module processes camera images in real time using image recognition algorithms (such as convolutional neural networks, CNNs) to identify stain types (the labeled training set includes features such as scale, silt, and algae), and evaluates the stain coverage area and concentration through pixel analysis. Data from all sensors is synchronized via timestamps, and a synchronous clock (such as IEEE 1588) ensures spatiotemporal consistency. Image data undergoes preprocessing such as denoising (median filtering) and grayscale conversion, while ranging data is noise-removed using Kalman filtering to generate structured data (such as point cloud maps and 2D raster maps).

[0036] Step S102. Divide the cleaning area according to the pool structure size data, combine the stain distribution image and stain type data, generate stain area marking results in the cleaning area, and determine the overall pollution level of the pool according to the water quality data, so as to retrieve the corresponding cleaning mode from the preset cleaning mode library. The preset cleaning mode library includes global coverage mode, key reinforcement mode, corner fitting mode and water quality maintenance mode. Among them, the global coverage mode is used to cover and clean at a uniform speed with a spiral path; the key reinforcement mode is used to clean areas with heavy stains by repetitive high-frequency brushing; the corner fitting mode is used to adjust the corresponding body angle and use the flexible brush head to fit and clean when dealing with the junction of the pool wall and bottom and the gap of the ladder; the water quality maintenance mode is used to release water purification agent and perform circulation filtration when the water quality data is abnormal.

[0037] Specifically, functional areas are divided based on structural data, cleaning strategies are matched by combining stain distribution and water quality data, and the optimal mode is retrieved from the preset mode library.

[0038] The pool is divided into sub-areas such as the pool bottom plane, vertical pool walls, corner junctions, ladder gaps, and curved walls; stain marking is done by marking areas with heavy / light stains on the area map (e.g., red marks areas with heavy algae, yellow marks areas with light sediment); pollution levels include: comprehensive water turbidity, stain coverage area and type, defining pollution levels (e.g., Level I light, Level II moderate, Level III heavy); pattern matching selects a single or combined pattern (e.g., global coverage + key enhancement) from the pattern library based on area characteristics and pollution levels.

[0039] The region segmentation algorithm identifies the boundary line between the pool wall and the pool bottom by using edge detection (Canny operator) based on ultrasonic point cloud data and camera images, and calculates the region's geometric parameters (such as the width of the escalator gap and the radius of curvature of the curved wall) by combining ranging data. The pool was divided into 50cm×50cm basic units using a grid map method, and the type (pool bottom, pool wall, corner, ladder) and structural features (such as curvature > 90° marked as arc-shaped area) of each unit were labeled.

[0040] Stain area marking and pollution level judgment are achieved by semantic segmentation of camera images (such as U-Net network), outputting the stain type and concentration value of each pixel, and setting a threshold (such as marking a pixel concentration >70% as a heavily stained area). The pollution level calculation formula includes: Pollution level = f(water turbidity, percentage of heavily soiled area, percentage of difficult-to-clean stains such as algae / scale); where f is a weighting function (e.g., water quality accounts for 30%, stain area accounts for 50%, and type accounts for 20%), and is divided into levels I-III.

[0041] Step S103. Based on the retrieved cleaning mode, generate the corresponding cleaning path plan and execution instructions to control the drive module, brushing module and purification module to work together.

[0042] Specifically, a precise path is generated based on the target pattern, and the driving, scrubbing, and purification modules are coordinated to perform tasks, ensuring efficient collaboration.

[0043] The path planning combines the results of area division and stain marking to generate a collision-free optimal path; through the parameter configuration of the drive module (motor speed, direction), the brushing module (frequency, pressure), and the purification module (release amount, timing), and avoids conflicts between modules (such as adjusting the path to avoid already released areas during purification).

[0044] The path planning algorithms include: Global Coverage Mode: Based on SLAM (Simultaneous Localization and Mapping) technology, a pool map is constructed. An improved Boustrophedon block division algorithm is used to divide the area into sub-blocks and generate a spiral coverage path. The overlap rate of adjacent paths is 10% to ensure no omissions. Focused Reinforcement Mode: For heavily soiled areas, a round-trip straight path is generated with a spacing equal to the width of the brush head (e.g., 20cm), and the number of coverages is ≥3. Corner Fitting Mode: The body tilt angle is calculated through a kinematic model (e.g., the angle between the pool wall and the pool bottom is θ, and the body tilt is θ-10°). The path moves along the arc of the boundary line, and the speed is reduced to 0.2m / s to ensure fit.

[0045] The hardware control strategy includes: Drive module: four-wheel differential drive, which adjusts the speed of the left and right wheels in real time according to the curvature of the path (e.g., the speed of the left wheel is less than that of the right wheel in an arc area), and is equipped with an IMU inertial measurement unit to calibrate the attitude in real time; Brushing module: servo motor drives the brush head, and in the key reinforcement mode, the torque is dynamically adjusted through current feedback (torque ≥ 0.5N*m), and the flexible brush head has a built-in pressure sensor, and the pressure is stable at 5-8N when in contact; Purification module: the storage tank has a capacity of 5L, and the release volume is controlled by a solenoid valve (e.g., 5mL of flocculant is released per square meter). After release, the circulation pump is started to work for 10 minutes, and at the same time, the robot is driven to move along the pool wall at a speed of 0.3m / s to promote diffusion.

[0046] Step S104. During the cleaning process, real-time status data of the cleaning area is acquired, and the cleaning area division and cleaning mode parameters are dynamically adjusted according to the real-time status data. If the current cleaning mode is found to be not performing up to the preset cleaning standard, a matching cleaning mode is retrieved from the cleaning mode library.

[0047] Specifically, by monitoring cleaning effectiveness through real-time data, the area division and mode parameters are dynamically optimized to form a closed-loop control.

[0048] Real-time status data: residual stain rate after cleaning, water quality improvement value, brush head fit (pressure feedback), motor energy consumption (overload detection); Adjustment mechanism: re-divide substandard areas, switch modes or correct parameters (such as increasing brushing frequency); Mode switching conditions: trigger re-retrieval logic when residual rate > 20% or water quality fails to meet standards. Effectiveness evaluation indicators: Stain removal rate: calculated by comparing before and after images (removal rate = (initial pixel concentration - current pixel concentration) / initial pixel concentration), preset standard ≥ 80%; Water quality compliance rate: turbidity < 20 NTU and pH value 6.5-8.5, if not met, extend the working time of the purification module; Fit feedback: when pressure sensor data < 3N, it is judged as insufficient fit, triggering a secondary adjustment of the body angle (±5°).

[0049] The cleaned area is rescanned every 5 minutes using a camera. If heavy stains (area > 0.5m²) are detected, the system will detect them. 2 Mark it as a new key area and insert it into the current path; determine the brush head load based on the motor current data (reduce the moving speed when overloaded and increase the brushing frequency when underloaded), and use the PID algorithm to dynamically adjust the speed and pressure; when the removal rate is detected to be <70% twice in a row, switch from "key enhancement mode" to "key enhancement + water quality maintenance joint mode" (simultaneously release detergent to assist brushing).

[0050] Abnormal handling includes: if the ultrasonic sensor detects an obstacle (such as debris suddenly entering the pool), cleaning will be immediately paused and an obstacle avoidance path will be initiated (returning to the breakpoint after bypassing the obstacle); when the battery level is less than 20%, it will automatically switch to "return mode" and return to the charging base along the shortest path.

[0051] In some embodiments, the real-time acquisition of water quality data, stain distribution images of the pool walls and bottom, pool structure size data, and stain type data via the equipped water quality sensor, camera, ultrasonic ranging module, and stain recognition module includes: controlling the water quality sensor to continuously acquire data on turbidity, pH value, residual chlorine concentration, and pollutant particle content of the pool water; controlling the camera to acquire images of the pool walls and bottom at a preset frame rate, generating continuous frame images containing the location and shape of stains; controlling the ultrasonic ranging module to emit ultrasonic signals to structures such as the pool walls, bottom, and ladders, calculating the distance data of each structure based on the echo time, and constructing a three-dimensional outline of the pool; and controlling the stain recognition module to identify the stain type as any one of mud, algae, scale, and grease based on image color features, texture features, and grayscale value differences, and marking the coordinate position of each stain type in the image.

[0052] By clearly defining the specific data collection targets, data types, and implementation methods of multiple sensors, the comprehensiveness and accuracy of environmental data can be ensured.

[0053] The water quality sensor focuses on four core indicators: turbidity, pH value, residual chlorine concentration, and pollutant particle content; the camera acquires images at a fixed frame rate, recording the spatial location and morphological characteristics of stains; ultrasonic ranging constructs a three-dimensional outline of the pool through echo time, covering structures such as walls, bottom, and ladders; stain recognition is based on color, texture, and grayscale differences, classifying stains into four typical types and locating them.

[0054] The water quality sensor control uses an integrated water quality sensor that collects data every 2 seconds to continuously monitor the water turbidity (unit NTU), pH value (accuracy ±0.01), residual chlorine concentration (mg / L), and pollutant particle content (measured by laser scattering method for the number of particles >5μm).

[0055] The camera is set to a frame rate of 25fps and a resolution of 1920×1080. It is equipped with a waterproof housing (IP68 rating) and the lens is coated with a dirt-resistant coating. When shooting, the robot's current position is used as the center, and a fisheye correction algorithm is used to generate a planar unfolded image, marking the dirt coordinates (x, y) and shape (block, sheet, flocculent).

[0056] By deploying six sets of ultrasonic sensors, distributed at the front, back, left, right and oblique angles of the body, emitting 40kHz sound waves every 100ms, the distance is calculated based on the echo time (formula: distance = speed of sound × time / 2); point cloud data of the pool wall, bottom and ladder are constructed by triangulation method to generate a three-dimensional contour model (accuracy ±1cm).

[0057] The stain recognition algorithm includes: color features: extracting hue (H) and saturation (S) features from the HSV color space (e.g., algae are green, H∈[40°, 80°]); texture features: calculating contrast and entropy values ​​through the gray-level co-occurrence matrix (GLCM) (e.g., scale has a rough texture, entropy value > 0.8); gray value difference: using threshold segmentation to distinguish stains from the background (e.g., mud stains have a gray value < 80, pool bottom background has a gray value > 150); the classifier trains the model using a support vector machine (SVM), inputs feature vectors and outputs stain type (mud, algae, scale, grease), and marks coordinates in the image (pixel-level positioning).

[0058] In some embodiments, before dividing the cleaning area according to the pool structure size data, the method further includes: performing noise reduction processing on the water quality data using a median filtering algorithm to remove sudden outliers; using a histogram equalization algorithm to enhance the contrast of the stain distribution image, improving the distinction between stains and the pool wall or bottom background; performing coordinate calibration on the pool structure size data obtained by the ultrasonic ranging module, establishing a local coordinate system with the initial position of the cleaning robot as the coordinate origin, and correcting measurement errors through multi-frame data fusion; and forming a standardized vector based on the preprocessed water quality data, stain distribution image, pool structure size data, and stain type data.

[0059] By denoising, enhancing, calibrating, and standardizing the raw sensor data, data quality is improved to support subsequent processing. Water quality data denoising removes burst noise through median filtering; image enhancement improves the contrast between stains and background through histogram equalization; coordinate calibration and data fusion establish a local coordinate system with the robot's initial position as the origin, and multi-frame data fusion corrects ranging errors; and vector standardization converts heterogeneous data into input vectors of a unified format.

[0060] Water quality data denoising involves using a 5-point median filter window on turbidity, pH, and other sequential data to remove outliers exceeding 3σ (such as jumps caused by sudden electromagnetic interference). Image contrast enhancement involves converting RGB images to grayscale and then applying a histogram equalization algorithm to expand the pixel grayscale value distribution to the full range of [0, 255], making the edges of stains clearer (e.g., increasing the grayscale difference between areas with mild algae and the pool wall from 20 to 50). Coordinate calibration and data fusion are achieved by establishing a local coordinate system: the x-axis points towards the long side of the pool, the y-axis points towards the short side, and the z-axis is vertically upward (the robot's initial position is (0, 0, 0)). Distance measurement data fusion uses a sliding window (10 frames) to apply Kalman filtering to the ultrasonic data, correcting measurement errors caused by water surface fluctuations (reducing the error from ±3cm to ±0.5cm).

[0061] Standardized vector construction involves converting water quality data (4-dimensional), image features (stain area, type one-hot encoding), and structural data (region coordinates, curvature) into a unified 128-dimensional feature vector, which is then normalized to the [0,1] interval (formula: (x-μ) / σ, where μ is the mean and σ is the standard deviation).

[0062] In some embodiments, the step of dividing the cleaning area according to the pool structure size data includes: dividing the pool into a bottom plane area, a vertical pool wall area, a circular arc area where the pool wall and bottom meet, a ladder area, and a handrail gap area based on the calibrated three-dimensional contour data of the pool; assigning a unique area identifier to each area, and setting the upper limit of the moving speed of the cleaning robot and the allowable range of the body tilt angle in the area according to the geometric characteristics of the area; for irregular pools, decomposing the complex structure into several calculable standard geometric units through a piecewise fitting algorithm, and dividing them into cleaning sub-areas respectively; the irregular pool includes curved pool walls and irregular steps.

[0063] Based on 3D contour data, the swimming pool is divided into functional sub-regions, regional characteristic parameters are set, and irregular structures are handled.

[0064] Standard area division: Divided into 5 types of areas: pool bottom, pool wall, boundary arc, ladder, and handrail gap; Define the upper limit of movement speed and the range of body tilt angle for each area; Irregular pool treatment decomposes complex structures into standard geometric units through segmented fitting.

[0065] Standard area division and marking include: Pool bottom plane area: a horizontal area with z-coordinate < 0.5m and curvature < 5°, marked R01, maximum movement speed 0.5m / s, allowable tilt angle range [-5°, 5°]; Vertical pool wall area: an area with z-coordinate 0.5-1.5m and normal vector perpendicular to the horizontal plane, marked R02, maximum speed 0.3m / s, allowable tilt angle range [0°, 20°] (for easy fitting); Intersecting arc area: an arc transition area where the angle between the pool wall and the pool bottom is < 135°, radius R ≤ 30cm, marked R03, maximum speed 0.2m / s, allowable tilt angle range [15°, 30°]; Escalator area: when a step-like structure is detected (height difference ≥ 15cm), marked R04, maximum speed 0.15m / s, high-speed movement is prohibited to prevent collision; Handrail gap area: a narrow area with a width < 10cm, marked R05, a flexible brush head must be used.

[0066] The irregular structure processing includes: for the arc-shaped pool wall (curvature radius R=50-100cm), piecewise Bézier curve fitting is used to decompose it into 5-10 arc segments (the central angle of each segment is ≤30°); for irregular steps (non-right-angle steps), the edge of the steps is fitted with straight lines, the height and width parameters of each step are defined, and independent sub-regions are generated (such as the step facade is marked as subclass R02).

[0067] In some embodiments, generating stain area marking results within the clean area by combining stain distribution images and stain type data includes: extracting stain contours from the preprocessed stain distribution image using an image segmentation algorithm, and calculating the area and average pixel grayscale value of each stain area; calculating a stain index for each stain area based on a preset stain degree weight for the stain type, combined with the stain area and average grayscale value; comparing the stain index with preset threshold ranges for heavy stain areas, moderate stain areas, and light stain areas, marking heavy stain areas, moderate stain areas, and light stain areas within the corresponding clean area, and attaching a stain type label to each marked area; wherein, the stain index corresponding to a heavy stain area is greater than a first threshold, the stain index corresponding to a moderate stain area is less than or equal to the first threshold and greater than a second threshold, the stain index corresponding to a light stain area is less than or equal to the second threshold, and the first threshold is greater than the second threshold.

[0068] By segmenting and weighting images, the degree of contamination in the stained areas is quantified, and different levels of areas are marked and labeled with type labels.

[0069] Stain contour extraction obtains stain boundaries based on image segmentation; stain index calculation combines stain type weights, area, and mean gray value; threshold comparison and labeling divide the area into heavy / moderate / lightly polluted regions and add type labels.

[0070] Stain contour extraction and feature calculation employ the U-Net semantic segmentation model (pre-trained on a swimming pool stain dataset), outputting a stain contour mask and calculating the area (unit: pixels). 2 Convert to actual area m 2 ) and the mean gray value (reflects the stain concentration, 0-255, the smaller the value, the higher the concentration).

[0071] The pollution index calculation formula includes: Pollution index = wt*T + wa*A / Amax + wg*(1-G / 255); where: wt is the type weight (scale = 0.4, algae = 0.3, silt = 0.2, grease = 0.4); wa and wg are the area and grayscale weights (both 0.3); the preset first threshold = 0.7 (severe), the second threshold = 0.4 (moderate), and <0.4 is mild; different colors are used to mark the area on the map: red (severe + scale), orange (severe + algae), yellow (moderate), and green (mild), and each marked area stores attributes such as coordinates, type, and pollution index.

[0072] In some embodiments, the preset cleaning standards include a residual stain area ratio of no more than 5% for lightly soiled areas, no more than 10% for moderately soiled areas, and no more than 20% for heavily soiled areas, and a decrease in key indicators in the water quality data of no less than 30%, including turbidity and pollutant particle content; if the current cleaning mode is found to be not meeting the preset cleaning standards, a matching cleaning mode is retrieved from the cleaning mode library, including: after the cleaning mode is completed or after cleaning each sub-area, comparing the real-time status data with the preset cleaning standards, if the residual area ratio or water quality indicator improvement of any type of stain area does not meet the standard, triggering the mode switching logic; wherein, the mode switching logic includes recalculating the cleaning priority and retrieving the matching cleaning mode based on the currently collected real-time water quality data, residual stain distribution image and pool structure data, and if the global coverage mode does not meet the standard, the key enhancement mode is retrieved to treat the residual area.

[0073] By defining residual standards and water quality improvement targets for different levels of stains and setting mode switching conditions, the cleaning effect is ensured to meet the standards.

[0074] The residual standards include: light ≤5%, moderate ≤10%, and heavy ≤20% residual area; water quality indicators include: turbidity and pollutant particle content decrease by ≥30%; the switching logic recalculates the priority and retrieves the matching mode (such as global to key enhancement) when the standards are not met.

[0075] After cleaning, each sub-area is photographed again, and the residual area percentage is calculated using the image difference method (formula: residual rate = number of residual pixels / number of initial stain pixels × 100%); the improvement of water quality indicators = (value before cleaning - value after cleaning) / value before cleaning × 100%, requiring an improvement of ≥30% in turbidity and particle content (e.g., turbidity reduced from 80 NTU to below 56 NTU).

[0076] The mode switching trigger conditions include: single area residual rate: severe > 20%, moderate > 10%, mild > 5%; global water quality improvement < 30%; priority treatment of the area with the highest residual rate. If the original mode is global coverage (not up to standard), the key enhancement mode is activated, and dense scrubbing paths are generated for the residual area (the number of round trips increases from 3 to 5).

[0077] In some embodiments, determining the overall pollution level of the swimming pool based on water quality data to retrieve the corresponding cleaning mode from a preset cleaning mode library includes: establishing a water pollution level judgment rule library, which defines multi-level threshold ranges for turbidity, pH value, residual chlorine concentration, and pollutant particle content; using a fuzzy logic algorithm to fuzzify the pollution degree of each water quality parameter to generate fuzzy membership values; weighting and summing the fuzzy membership values ​​of each parameter using preset weighting coefficients to obtain the overall pollution level of the swimming pool; when the overall pollution level reaches a preset activation threshold, prioritizing the retrieval of the water quality maintenance mode and adjusting the purifying agent release amount and circulation filtration time according to the pollution level.

[0078] The overall pollution level is calculated by integrating multiple water quality parameters using a fuzzy logic algorithm, which drives water quality maintenance modes and parameter adjustments. A rule base is established by defining multi-level threshold ranges for each parameter; fuzzification converts parameters into "low, medium, and high" membership values; and weighted summation yields the pollution level, which is then used to adjust the purification agent release strategy.

[0079] The water pollution level rule base is shown in the table below:

[0080] Fuzzy logic calculation calculates the membership degree for each parameter (e.g., turbidity = 60 NTU, membership degree for "high pollution" = 0.8, "medium pollution" = 0.2); the weighted formula includes: pollution level = 0.4 × turbidity membership degree + 0.3 × particle membership degree + 0.2 × pH membership degree + 0.1 × residual chlorine membership degree, the result ∈ [0,1], ≥0.6 activates the water quality maintenance mode.

[0081] The purification strategy adjustments include: for light pollution (0.4-0.6): release 50 mL of flocculant and circulate for 15 minutes; for heavy pollution (>0.6): release 100 mL of special algaecide and circulate for 30 minutes, while reducing the scrubbing speed to prolong the action time of the purifying agent.

[0082] In some embodiments, generating corresponding cleaning path planning and execution instructions based on the retrieved cleaning mode, and controlling the drive module, brushing module, and purification module to work collaboratively, includes: if the global coverage mode is retrieved, generating a spiral coverage path starting from the current position of the cleaning robot based on the coordinate range of the bottom plane area, with the path spacing not exceeding the effective cleaning width of the brushing module, and controlling the drive module to move at a constant speed; if the focused reinforcement mode is retrieved, for the marked heavily soiled areas, generating a reciprocating cleaning sub-path by extending 10 centimeters outward from the area boundary coordinates, and controlling the brush head of the brushing module. The rotation speed is increased to a high-frequency setting, increasing the contact pressure between the brush head and the dirt surface. If the corner fitting mode is selected, the tilt angle of the cleaning robot body is calculated based on the radius of curvature of the arc area where the pool wall and bottom meet, so that the brush head axis is perpendicular to the tangent of the arc. The flexible brush head is then switched, and the drive module is controlled to move along the junction line at a low speed. If the water quality maintenance mode is selected, the corresponding type of purifier is released from the purifier storage compartment according to the type of pollutant in the water quality data. The built-in filter pump is activated to circulate and filter the water at a flow rate 1.5 times that of the pool water circulation flow rate, and the uniformity of the purifier diffusion is monitored in real time by a camera.

[0083] For different cleaning modes, it generates exclusive path planning and hardware control commands to achieve multi-module collaboration. Global coverage includes: spiral path, uniform speed movement; key reinforcement includes: reciprocating path with extended boundaries, high-frequency and high-pressure brushing; corner fitting includes: tilting the body to fit the arc, flexible brush head moving at low speed; water quality maintenance includes: releasing corresponding purifying agents and controlling the filter pump flow.

[0084] The global coverage mode includes: Path planning: Starting from the current position, a spiral path with a pitch of 10cm = brush head width - 10cm is generated (e.g., brush head width 30cm, pitch 20cm) to ensure a 10% overlap rate; Drive control: Four-wheel differential drive, speed 0.4m / s, IMU real-time calibration of heading angle (automatic correction when deviation > 2°).

[0085] Key enhancement modes include: Path expansion: Expanding the coordinates (x1, y1, x2, y2) of heavily soiled areas by 10cm (x1-10cm, y1-10cm, x2+10cm, y2+10cm) to generate a round-trip straight path (spacing = brush head width); Brush control: Increasing the brush head speed from 50rpm to 120rpm, and adjusting the brush head pressure to 8-10N via pressure sensor feedback (increasing the torque of the drive motor when the contact force is insufficient).

[0086] The corner fitting mode includes: Angle calculation: Based on the radius of curvature R of the boundary arc, the body tilt angle θ = arctan(brush head radius / R) (e.g., R = 20cm, brush head radius 5cm, θ≈14°), ensuring that the brush head axis is perpendicular to the tangent of the arc; Hardware switching: Enable the silicone flexible brush head (Shore A 40° hardness), reduce the drive speed to 0.15m / s, and perform circular interpolation motion along the boundary line.

[0087] Water quality maintenance modes include: Purifier release: The release amount is controlled by a solenoid valve according to the type of pollutant (e.g., copper sulfate released by algae, citric acid released by scale) (release amount per square meter = pollution level × 20 mL); Filtration control: A circulation pump with a flow rate of 150 L / h is started (pool water circulation speed 1.5 times), and a camera monitors the diffusion of purifier in real time (circulation stops when color uniformity > 90%).

[0088] In some embodiments, the step of dynamically adjusting the cleaning area division and cleaning mode parameters based on real-time status data includes: acquiring images of the cleaned area in real time using a camera, calculating the difference between the images and the images of the stain distribution before cleaning, and identifying the location of any remaining stains; if the area of ​​the remaining stains accounts for more than 15% of the area of ​​the original marked area, remarking the remaining area corresponding to the location of the remaining stains as a new medium-stain area, and adding a corresponding sub-area to the existing cleaning area division; and dynamically adjusting the moving speed and brush head pressure of the cleaning robot in complex structural areas based on the real-time transmitted current value of the drive module and the torque value of the brush head motor, wherein when the current value exceeds 120% of the rated current, the moving speed is automatically reduced by 50% and the brush head pressure is reduced by 30%.

[0089] Residual stains are identified through real-time image comparison, the area division is dynamically updated, and cleaning parameters are adjusted based on hardware feedback. Residual stain identification includes: difference calculation to locate uncleaned areas; area remarking includes: marking areas with more than 15% residue as new moderate areas; parameter adaptation includes: dynamically adjusting speed and pressure based on current and torque to avoid overload.

[0090] Residual stain detection includes: performing grayscale difference analysis between the cleaned image and the preprocessed original image (formula: ΔG=|G after cleaning - G before cleaning|), and determining the area of ​​residual stains as ΔG<30, and calculating their area percentage.

[0091] The region re-division logic includes: if the residual percentage is >15%, it is marked as a new region on the map (e.g., the original heavily polluted region is marked as a moderate region R06 after the residual is removed), and inserted into the current cleaning queue with a higher priority than the uncleaned lightly polluted regions; when regenerating the path, the shortest path to the new region is planned using the A* algorithm to avoid repeatedly covering regions that have already met the standards.

[0092] The drive module current monitoring includes: when the rated current is 1.5A and exceeds 120% (1.8A), it is determined that the brush head is stuck (such as stuck in the gap of the escalator), the speed drops from 0.3m / s to 0.15m / s, and the brush head pressure drops from 10N to 7N; the brush head motor torque feedback includes: real-time monitoring by the torque sensor, when underloaded (torque < 0.3N*m), the speed is increased to 150rpm to make full use of the cleaning capacity.

[0093] In some embodiments, an environment perception framework based on multimodal Transformer is proposed, which integrates visual, ultrasonic, and water quality data for joint modeling, and solves the problem of insufficient stain sample labeling through self-supervised learning, thereby achieving robust stain detection and classification.

[0094] Multimodal data encoding includes: a visual branch: using an improved YOLOv8 model, embedding a focus attention mechanism (CBAM) to focus on the stained area, and outputting the stained location mask and type probability (silt / algae, etc.); an ultrasonic branch: encoding 3D point cloud data into structural feature vectors (including geometric properties such as curvature and roughness) through PointNet++; a water quality branch: mapping parameters such as turbidity and residual chlorine into water quality feature vectors through a fully connected layer; and a fusion layer that learns the correlation between sensors (such as the co-occurrence probability of high turbidity areas and silt stains) through the cross-modal attention mechanism of Transformer.

[0095] The self-supervised pre-training strategy includes: masked image modeling: randomly occluding 30% of the image area, training the model to predict the type and texture of the stain in the missing part (using a contrastive loss function); ultrasonic data augmentation: adding Gaussian noise to the 3D point cloud, and learning the invariant representation of structural features through a Siamese network; pseudo-label generation: automatically labeling unlabeled data using a thresholding method (e.g., areas with gray values ​​< 50 and ultrasonic roughness > 0.8 are labeled as "severe mud and sand stains"), and constructing a weakly supervised training set.

[0096] Real-time inference optimization optimizes the inference latency to within 50ms by deploying the model to an edge computing module (such as NVIDIA Jetson AGX Orin) and using model quantization technology. The fused output includes stain coordinates (accuracy ±5cm), type confidence (>0.7 triggers an alarm), and structural risk level (such as the collision risk index of escalator gaps).

[0097] In some embodiments, by constructing a graph neural network (GNN) model, the pool structure is abstracted into graph nodes with attributes. The correlation between regions is learned through graph convolution, realizing intelligent region division based on the law of pollution propagation, and using a spatiotemporal prediction model to predict the trend of stain diffusion.

[0098] The pool structure diagram modeling includes: Node definition: Each standard area (such as pool bottom partition, pool wall panel) is a graph node, and the attributes include geometric parameters (curvature, area) and historical pollution data (the frequency of algal blooms in the past 7 days); Edge definition: The connection relationship between nodes (such as the water flow resistance coefficient of the boundary between the pool bottom and the pool wall), and the weight reflects the probability of pollutant migration (such as when the water flow velocity is >0.2m / s, the edge weight is +0.3).

[0099] Pollution prediction and regional dynamic division include: Spatiotemporal graph convolutional network (ST-GCN): Input water quality data (time dimension) and regional pollution index (spatial dimension) for the past 24 hours, predict the pollution index change trend of each region in the next 2 hours (e.g., due to water stagnation in the escalator area, the algae growth rate is predicted to increase by 15%); Intelligent division strategy: If the pollution index of a certain pool wall area is predicted to exceed the threshold, it will be automatically upgraded from "moderate area" to "key monitoring area", and a pre-cleaning path will be planned in advance.

[0100] The cleaning priority ranking algorithm is based on the shortest path first algorithm of the graph (an improvement on the Dijkstra algorithm): combined with the predicted pollution spread rate, it prioritizes the processing of "high-risk nodes" (such as the boundary arc nodes connecting multiple regions, whose pollution spread impact is calculated by the PageRank algorithm); after each sub-region is cleaned, the pollution status of the graph nodes is updated, triggering the GNN to recalculate the regional correlation (such as the pollution migration probability of adjacent regions decreases after a certain region is cleaned).

[0101] In some embodiments, by constructing a digital twin model of pool cleaning, the physical environment and robot status are mapped in real time, and the cleaning strategy is dynamically updated by combining online learning algorithms to achieve an intelligent closed loop of "prediction-execution-feedback".

[0102] The construction of the digital twin includes: Geometric twin: generating a high-precision 3D model of the swimming pool based on ultrasonic point cloud data (accuracy ±0.5cm), integrating a water flow simulation module (CFD simplified model, calculating the water flow velocity field in different areas); Behavioral twin: establishing a robot dynamics model (considering water pressure resistance and brush head friction), inputting drive commands to predict the movement trajectory (position error <2cm); Pollution twin: simulating the diffusion process of purifying agents through reaction-diffusion equations (such as predicting the concentration distribution of copper sulfate in pool water, with an error <8%).

[0103] The online learning optimization mechanism includes: Model update: After each cleaning task is completed, the actual cleaning effect (residual rate, water quality improvement data) is compared with the digital twin prediction results, and the pollution diffusion model parameters are updated using an online gradient descent algorithm; Strategy iteration: If the actual residual rate is more than 10% higher than the predicted value, a strategy correction is triggered: the brushing pressure for this type of stain is increased by 2N, the brush head speed is increased by 10rpm, and the effect of the adjustment is verified in the digital twin; Anomaly detection: By identifying the state difference between the twin and the physical robot (such as the difference in drive motor current > 15%), equipment malfunctions (such as brush head jamming) are identified, and the system automatically switches to the backup cleaning module.

[0104] Multi-agent collaborative extension extends the pool cleaning data shared by multiple robots through federated learning technology when multiple robots are deployed (only encrypted model parameters are uploaded to protect user privacy); the central server periodically aggregates global knowledge and updates general cleaning strategies (such as automatically increasing the default value of algae removal agent release if a high frequency of algae blooms is detected in a certain area during the rainy season).

[0105] This application also provides a cleaning mode switching device for a multi-functional swimming pool cleaning robot. This cleaning mode switching device is used to execute the steps of the cleaning mode switching method for the multi-functional swimming pool cleaning robot shown in the above embodiments. The cleaning mode switching device can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, a laptop computer, a wearable device, or a robot.

[0106] The cleaning mode switching device for the multi-functional pool cleaning robot includes: The data acquisition unit is used to collect real-time water quality data, images of stain distribution on the pool walls and bottom, pool structural dimensions, and stain type data of the swimming pool through the built-in water quality sensor, camera, ultrasonic ranging module, and stain recognition module. The pattern acquisition unit is used to divide the cleaning area according to the pool's structural dimensions, combine the stain distribution image and stain type data, generate stain area marking results within the cleaning area, and determine the overall pollution level of the pool based on water quality data. It then retrieves the corresponding cleaning mode from a preset cleaning mode library, which includes a global coverage mode, a focused cleaning mode, a corner-fitting mode, and a water quality maintenance mode. The global coverage mode is used for uniform coverage cleaning along a spiral path; the focused cleaning mode is used for cleaning heavily soiled areas with repetitive high-frequency brushing; the corner-fitting mode is used to adjust the machine angle and activate a flexible brush head for cleaning the junction of the pool wall and bottom and the gaps in the ladder; and the water quality maintenance mode is used to release a water purifier and perform circulating filtration when water quality data is abnormal. The instruction execution unit is used to generate corresponding cleaning path planning and execution instructions based on the retrieved cleaning mode, and control the drive module, brushing module and purification module to work together. The mode switching unit is used to acquire real-time status data of the cleaning area during the cleaning process, dynamically adjust the cleaning area division and cleaning mode parameters based on the real-time status data, and if it is detected that the current cleaning mode does not meet the preset cleaning standard, it retrieves a matching cleaning mode from the cleaning mode library.

[0107] It should be noted that those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the cleaning mode switching device and each module of the multi-functional pool cleaning robot described above can be referred to the corresponding process in the embodiments of the cleaning mode switching method of the multi-functional pool cleaning robot described above, and will not be repeated here.

[0108] The cleaning mode switching method of the aforementioned multi-functional pool cleaning robot can be implemented as a computer program that can run on the device provided in this application.

[0109] This application also provides a cleaning robot. The cleaning robot includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0110] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any cleaning mode switching method for the multi-functional pool cleaning robot.

[0111] The processor provides computing and control capabilities to support the operation of the entire cleaning robot.

[0112] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any cleaning mode switching method of the multi-functional pool cleaning robot.

[0113] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that the structures shown in the above embodiments are only partial structures related to the present application and do not constitute a limitation on the terminal to which the present application is applied. Specific cleaning robots may include more or fewer components than mentioned in the embodiments, or combine certain components, or have different component arrangements.

[0114] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0115] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: The system uses a water quality sensor, camera, ultrasonic ranging module, and stain recognition module to collect real-time data on the pool's water quality, stain distribution on the pool walls and bottom, pool structure dimensions, and stain types. The cleaning area is divided based on the pool's structural dimensions. Combined with stain distribution images and stain type data, stain area markings are generated within each cleaning area. The overall pollution level of the pool is determined based on water quality data, and a corresponding cleaning mode is retrieved from a pre-set cleaning mode library. This library includes a global coverage mode, a focused cleaning mode, a corner-fitting mode, and a water quality maintenance mode. The global coverage mode uses a spiral path for uniform cleaning. The focused cleaning mode uses a repetitive, high-frequency brushing method to clean heavily soiled areas. The corner-fitting mode adjusts the machine angle and uses a flexible brush head for cleaning the junction of the pool wall and bottom, and the gaps in the ladder. The water quality maintenance mode releases a water purifier and performs circulating filtration when water quality data is abnormal. Based on the retrieved cleaning mode, the corresponding cleaning path plan and execution instructions are generated to control the drive module, scrubbing module and purification module to work together. During the cleaning process, real-time status data of the cleaning area is acquired, and the cleaning area division and cleaning mode parameters are dynamically adjusted based on the real-time status data. If the current cleaning mode is found to be not performing up to the preset cleaning standard, a matching cleaning mode is retrieved from the cleaning mode library.

[0116] In some embodiments, the real-time acquisition of water quality data, stain distribution images of the pool walls and bottom, pool structure size data, and stain type data via the equipped water quality sensor, camera, ultrasonic ranging module, and stain recognition module includes: controlling the water quality sensor to continuously acquire data on turbidity, pH value, residual chlorine concentration, and pollutant particle content of the pool water; controlling the camera to acquire images of the pool walls and bottom at a preset frame rate, generating continuous frame images containing the location and shape of stains; controlling the ultrasonic ranging module to emit ultrasonic signals to structures such as the pool walls, bottom, and ladders, calculating the distance data of each structure based on the echo time, and constructing a three-dimensional outline of the pool; and controlling the stain recognition module to identify the stain type as any one of mud, algae, scale, and grease based on image color features, texture features, and grayscale value differences, and marking the coordinate position of each stain type in the image.

[0117] In some embodiments, before dividing the cleaning area according to the pool structure size data, the method further includes: performing noise reduction processing on the water quality data using a median filtering algorithm to remove sudden outliers; using a histogram equalization algorithm to enhance the contrast of the stain distribution image, improving the distinction between stains and the pool wall or bottom background; performing coordinate calibration on the pool structure size data obtained by the ultrasonic ranging module, establishing a local coordinate system with the initial position of the cleaning robot as the coordinate origin, and correcting measurement errors through multi-frame data fusion; and forming a standardized vector based on the preprocessed water quality data, stain distribution image, pool structure size data, and stain type data.

[0118] In some embodiments, the step of dividing the cleaning area according to the pool structure size data includes: dividing the pool into a bottom plane area, a vertical pool wall area, a circular arc area where the pool wall and bottom meet, a ladder area, and a handrail gap area based on the calibrated three-dimensional contour data of the pool; assigning a unique area identifier to each area, and setting the upper limit of the moving speed of the cleaning robot and the allowable range of the body tilt angle in the area according to the geometric characteristics of the area; for irregular pools, decomposing the complex structure into several calculable standard geometric units through a piecewise fitting algorithm, and dividing them into cleaning sub-areas respectively; the irregular pool includes curved pool walls and irregular steps.

[0119] In some embodiments, generating stain area marking results within the clean area by combining stain distribution images and stain type data includes: extracting stain contours from the preprocessed stain distribution image using an image segmentation algorithm, and calculating the area and average pixel grayscale value of each stain area; calculating a stain index for each stain area based on a preset stain degree weight for the stain type, combined with the stain area and average grayscale value; comparing the stain index with preset threshold ranges for heavy stain areas, moderate stain areas, and light stain areas, marking heavy stain areas, moderate stain areas, and light stain areas within the corresponding clean area, and attaching a stain type label to each marked area; wherein, the stain index corresponding to a heavy stain area is greater than a first threshold, the stain index corresponding to a moderate stain area is less than or equal to the first threshold and greater than a second threshold, the stain index corresponding to a light stain area is less than or equal to the second threshold, and the first threshold is greater than the second threshold.

[0120] In some embodiments, the preset cleaning standards include a residual stain area ratio of no more than 5% for lightly soiled areas, no more than 10% for moderately soiled areas, and no more than 20% for heavily soiled areas, and a decrease in key indicators in the water quality data of no less than 30%, including turbidity and pollutant particle content; if the current cleaning mode is found to be not meeting the preset cleaning standards, a matching cleaning mode is retrieved from the cleaning mode library, including: after the cleaning mode is completed or after cleaning each sub-area, comparing the real-time status data with the preset cleaning standards, if the residual area ratio or water quality indicator improvement of any type of stain area does not meet the standard, triggering the mode switching logic; wherein, the mode switching logic includes recalculating the cleaning priority and retrieving the matching cleaning mode based on the currently collected real-time water quality data, residual stain distribution image and pool structure data, and if the global coverage mode does not meet the standard, the key enhancement mode is retrieved to treat the residual area.

[0121] In some embodiments, determining the overall pollution level of the swimming pool based on water quality data to retrieve the corresponding cleaning mode from a preset cleaning mode library includes: establishing a water pollution level judgment rule library, which defines multi-level threshold ranges for turbidity, pH value, residual chlorine concentration, and pollutant particle content; using a fuzzy logic algorithm to fuzzify the pollution degree of each water quality parameter to generate fuzzy membership values; weighting and summing the fuzzy membership values ​​of each parameter using preset weighting coefficients to obtain the overall pollution level of the swimming pool; when the overall pollution level reaches a preset activation threshold, prioritizing the retrieval of the water quality maintenance mode and adjusting the purifying agent release amount and circulation filtration time according to the pollution level.

[0122] In some embodiments, generating corresponding cleaning path planning and execution instructions based on the retrieved cleaning mode, and controlling the drive module, brushing module, and purification module to work collaboratively, includes: if the global coverage mode is retrieved, generating a spiral coverage path starting from the current position of the cleaning robot based on the coordinate range of the bottom plane area, with the path spacing not exceeding the effective cleaning width of the brushing module, and controlling the drive module to move at a constant speed; if the focused reinforcement mode is retrieved, for the marked heavily soiled areas, generating a reciprocating cleaning sub-path by extending 10 centimeters outward from the area boundary coordinates, and controlling the brush head of the brushing module. The rotation speed is increased to a high-frequency setting, increasing the contact pressure between the brush head and the dirt surface. If the corner fitting mode is selected, the tilt angle of the cleaning robot body is calculated based on the radius of curvature of the arc area where the pool wall and bottom meet, so that the brush head axis is perpendicular to the tangent of the arc. The flexible brush head is then switched, and the drive module is controlled to move along the junction line at a low speed. If the water quality maintenance mode is selected, the corresponding type of purifier is released from the purifier storage compartment according to the type of pollutant in the water quality data. The built-in filter pump is activated to circulate and filter the water at a flow rate 1.5 times that of the pool water circulation flow rate, and the uniformity of the purifier diffusion is monitored in real time by a camera.

[0123] In some embodiments, the step of dynamically adjusting the cleaning area division and cleaning mode parameters based on real-time status data includes: acquiring images of the cleaned area in real time using a camera, calculating the difference between the images and the images of the stain distribution before cleaning, and identifying the location of any remaining stains; if the area of ​​the remaining stains accounts for more than 15% of the area of ​​the original marked area, remarking the remaining area corresponding to the location of the remaining stains as a new medium-stain area, and adding a corresponding sub-area to the existing cleaning area division; and dynamically adjusting the moving speed and brush head pressure of the cleaning robot in complex structural areas based on the real-time transmitted current value of the drive module and the torque value of the brush head motor, wherein when the current value exceeds 120% of the rated current, the moving speed is automatically reduced by 50% and the brush head pressure is reduced by 30%.

[0124] The embodiments of this application also provide a computer-readable storage medium storing a computer program, the computer program including program instructions, and the processor executing the program instructions to implement the steps of the cleaning mode switching method of the multi-functional cleaning robot for swimming pools provided in the above embodiments of this application.

[0125] The computer-readable storage medium can be the internal storage unit of the cleaning robot described in the foregoing embodiments, such as the hard drive or memory of the cleaning robot. Alternatively, the computer-readable storage medium can be an external storage device of the cleaning robot, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the cleaning robot.

[0126] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for switching cleaning modes of a multi-functional swimming pool cleaning robot, characterized in that, Applications in cleaning robots, including: The system uses a water quality sensor, camera, ultrasonic ranging module, and stain recognition module to collect real-time data on the pool's water quality, stain distribution on the pool walls and bottom, pool structure dimensions, and stain types. The cleaning area is divided based on the pool's structural dimensions. Combined with stain distribution images and stain type data, stain area markings are generated within each cleaning area. The overall pollution level of the pool is determined based on water quality data, and a corresponding cleaning mode is retrieved from a pre-set cleaning mode library. This library includes a global coverage mode, a focused cleaning mode, a corner-fitting mode, and a water quality maintenance mode. The global coverage mode uses a spiral path for uniform cleaning. The focused cleaning mode uses a repetitive, high-frequency brushing method to clean heavily soiled areas. The corner-fitting mode adjusts the machine angle and uses a flexible brush head for cleaning the junction of the pool wall and bottom, and the gaps in the ladder. The water quality maintenance mode releases a water purifier and performs circulating filtration when water quality data is abnormal. Based on the retrieved cleaning mode, the corresponding cleaning path plan and execution instructions are generated to control the drive module, scrubbing module and purification module to work together. During the cleaning process, real-time status data of the cleaning area is acquired, and the cleaning area division and cleaning mode parameters are dynamically adjusted based on the real-time status data. If the current cleaning mode is found to be not performing up to the preset cleaning standard, a matching cleaning mode is retrieved from the cleaning mode library.

2. The method according to claim 1, characterized in that, The system utilizes a water quality sensor, camera, ultrasonic ranging module, and stain recognition module to collect real-time data on the pool's water quality, stain distribution images on the pool walls and bottom, pool structural dimensions, and stain type, including: The water quality sensor continuously collects data on turbidity, pH value, residual chlorine concentration, and pollutant particle content in the pool water. The camera is controlled to capture images of the pool wall and bottom at a preset frame rate, generating continuous frame images that include the location and shape of the stains; The ultrasonic ranging module is controlled to emit ultrasonic signals to the pool wall, bottom and ladder structures, and the distance data of each structure is calculated based on the echo time to construct the three-dimensional outline of the pool. The stain recognition module identifies stain types as mud, algae, scale, and grease based on image color features, texture features, and grayscale differences, and marks the coordinate position of each stain type in the image.

3. The method according to claim 1, characterized in that, Before dividing the cleaning area according to the pool structure size data, the method also includes: The water quality data is denoised by using a median filtering algorithm to remove sudden outliers. Histogram equalization algorithm is used to enhance the contrast of the stain distribution image, thereby improving the distinction between the stain and the background of the pool wall or bottom. The pool structure dimension data acquired by the ultrasonic ranging module is calibrated by coordinates. A local coordinate system is established with the initial position of the cleaning robot as the coordinate origin, and the measurement error is corrected by fusing multiple frames of data. A standardized vector is generated based on the pretreated water quality data, stain distribution images, pool structure size data, and stain type data.

4. The method according to claim 1, characterized in that, The process of dividing the cleaning area according to the pool's structural dimensions includes: Based on the calibrated 3D contour data of the pool, the pool is divided into the bottom plane area, the vertical pool wall area, the arc area where the pool wall and the bottom meet, the ladder area, and the handrail gap area. Each area is assigned a unique area identifier, and the maximum speed limit for the cleaning robot and the allowable range of body tilt angle in that area are set according to the geometric characteristics of the area. For irregular swimming pools, the complex structure is decomposed into several computable standard geometric units using a piecewise fitting algorithm, and each unit is divided into a clean sub-region; the irregular swimming pool includes curved pool walls and irregularly shaped steps.

5. The method according to claim 1, characterized in that, The process of combining stain distribution images and stain type data to generate stain area marking results within the cleaning area includes: The stain contours are extracted from the preprocessed stain distribution image using an image segmentation algorithm, and the area and average pixel gray value of each stain region are calculated. Based on the preset pollution degree weight of the stain type, and combined with the stain area and the mean gray value, the pollution index of each stain area is calculated. The contamination index is compared with preset threshold ranges for heavily stained areas, moderately stained areas, and lightly stained areas. The corresponding cleaning areas are marked as heavily stained areas, moderately stained areas, and lightly stained areas, and a stain type label is attached to each marked area. The contamination index corresponding to a heavily stained area is greater than a first threshold, the contamination index corresponding to a moderately stained area is less than or equal to the first threshold and greater than a second threshold, and the contamination index corresponding to a lightly stained area is less than or equal to the second threshold. The first threshold is greater than the second threshold.

6. The method according to claim 5, characterized in that, The preset cleaning standards include that the residual stain area in lightly soiled areas does not exceed 5%, in moderately soiled areas it does not exceed 10%, and in heavily soiled areas it does not exceed 20%, and the key indicators in the water quality data decrease by no less than 30%. Key indicators include turbidity and pollutant particle content. If the current cleaning mode is found to be not meeting the preset cleaning standards, a matching cleaning mode is retrieved from the cleaning mode library, including: After the cleaning mode is completed or after each sub-area is cleaned, the real-time status data is compared with the preset cleaning standards. If the residual area ratio of any type of stain or the improvement of water quality indicators does not meet the standard, the mode switching logic is triggered. The mode switching logic includes recalculating the cleaning priority and retrieving the matching cleaning mode based on the currently collected water quality data, residual stain distribution images, and pool structure data. If the global coverage mode fails to meet the requirements, the key enhancement mode is retrieved to treat the residual areas.

7. The method according to claim 1, characterized in that, The step of determining the overall pollution level of the swimming pool based on water quality data and retrieving the corresponding cleaning mode from a preset cleaning mode library includes: Establish a rule base for judging water pollution levels, defining multi-level threshold ranges for turbidity, pH value, residual chlorine concentration, and pollutant particulate content in the rule base; The pollution degree of each water quality parameter is fuzzified using a fuzzy logic algorithm to generate fuzzy membership values; The overall pollution level of the swimming pool is obtained by weighting and summing the fuzzy membership degrees of each parameter using preset weighting coefficients. When the overall pollution level reaches the preset activation threshold, the water quality maintenance mode is activated first, and the amount of purifying agent released and the circulation filtration time are adjusted according to the pollution level.

8. The method according to claim 1, characterized in that, The process of generating corresponding cleaning path planning and execution instructions based on the retrieved cleaning mode, and controlling the collaborative work of the drive module, scrubbing module, and purification module includes: If the global coverage mode is invoked, a spiral coverage path is generated starting from the current position of the cleaning robot based on the coordinate range of the bottom plane area. The path spacing does not exceed the effective cleaning width of the brushing module, and the drive module is controlled to move at a constant speed. If the focus enhancement mode is activated, for the marked heavily soiled areas, a reciprocating cleaning sub-path is generated by extending 10 centimeters outward from the area boundary coordinates, the brush head speed of the scrubbing module is increased to a high frequency level, and the contact pressure between the brush head and the soiled surface is increased. If the corner fitting mode is selected, the tilt angle of the cleaning robot body is calculated based on the radius of curvature of the arc area where the pool wall and the pool bottom meet, so that the brush head axis is perpendicular to the tangent of the arc, and the flexible brush head is switched to control the drive module to move along the junction line at a low speed. If the water quality maintenance mode is activated, the corresponding type of purifier will be released from the purifier storage compartment according to the type of pollutants in the water quality data, and the built-in filter pump will be started to circulate and filter the water at a flow rate 1.5 times that of the pool water circulation flow rate. The uniformity of the purifier diffusion will be monitored in real time by a camera.

9. The method according to claim 1, characterized in that, The method of dynamically adjusting the cleaning area division and cleaning mode parameters based on real-time status data includes: The system uses a camera to capture images of the cleaned area in real time, and calculates the difference between these images and the images of the stain distribution before cleaning to identify the location of any remaining stains. If the area of ​​residual stains exceeds 15% of the area of ​​the original marked area, the residual area corresponding to the location of the residual stains will be remarked as a new medium stain area, and a corresponding sub-area will be added to the existing cleaning area division. Based on the real-time feedback of the drive module current value and the brush head motor torque value, the cleaning robot's moving speed and brush head pressure are dynamically adjusted in complex structural areas. When the current value exceeds 120% of the rated current, the moving speed is automatically reduced by 50% and the brush head pressure is reduced by 30%.

10. A cleaning mode switching device for a multi-functional swimming pool cleaning robot, characterized in that, Applications in cleaning robots, including: The data acquisition unit is used to collect real-time water quality data, images of stain distribution on the pool walls and bottom, pool structural dimensions, and stain type data of the swimming pool through the built-in water quality sensor, camera, ultrasonic ranging module, and stain recognition module. The pattern acquisition unit is used to divide the cleaning area according to the pool's structural dimensions, combine the stain distribution image and stain type data, generate stain area marking results within the cleaning area, and determine the overall pollution level of the pool based on water quality data. It then retrieves the corresponding cleaning mode from a preset cleaning mode library, which includes a global coverage mode, a focused cleaning mode, a corner-fitting mode, and a water quality maintenance mode. The global coverage mode is used for uniform coverage cleaning along a spiral path; the focused cleaning mode is used for cleaning heavily soiled areas with repetitive high-frequency brushing; the corner-fitting mode is used to adjust the machine angle and activate a flexible brush head for cleaning the junction of the pool wall and bottom and the gaps in the ladder; and the water quality maintenance mode is used to release a water purifier and perform circulating filtration when water quality data is abnormal. The instruction execution unit is used to generate corresponding cleaning path planning and execution instructions based on the retrieved cleaning mode, and control the drive module, brushing module and purification module to work together. The mode switching unit is used to acquire real-time status data of the cleaning area during the cleaning process, dynamically adjust the cleaning area division and cleaning mode parameters based on the real-time status data, and if it is detected that the current cleaning mode does not meet the preset cleaning standard, it retrieves a matching cleaning mode from the cleaning mode library.

Images