Secondary water supply tank cleaning robot intelligent control method and system

Abstract

The application discloses a secondary water supply water tank cleaning robot intelligent control method and system, relates to the technical field of water tank cleaning, and spatial modeling is performed on the water body state in the water tank through a sediment perception and data acquisition module. The system can identify the specific area where the sediment is accumulated, and no longer relies on the blind traversal mode of the traditional fixed path, so that the cleaning behavior is changed from overall expansion to focused attention. The dynamic determination logic of the sediment hot spot index and the cleaning priority level is introduced, the area with high cleaning difficulty and dense sediment in the water tank can be actively identified, the area is preferentially included in the robot path for execution and processing, and the pertinence and effect coverage of the cleaning action are improved. Through the data archiving and optimization learning module, the system stores and learns the sediment distribution characteristics, trajectory execution records and coverage rate data formed in each cleaning process, extracts inefficient areas and task performance bottlenecks, and gradually optimizes the operation efficiency and cleaning effect.

Description

Technical Field

[0001] This invention relates to the field of water supply tank cleaning technology, specifically to an intelligent control method and system for a secondary water supply tank cleaning robot. Background Technology

[0002] In urban water supply systems, water quality assurance relies not only on primary water treatment facilities, but also on secondary water supply tanks, which serve as transitional regulators and end-point safeguards. With increasing building density and the decentralization of user terminals, traditional water tank systems, which are manually cleaned periodically, are gradually being replaced by cleaning robots. This invention addresses this type of robot control system for "removing deposits from the internal walls of secondary water supply tanks."

[0003] In practical applications, most existing cleaning robots employ static path planning. This means that a fixed trajectory is set before cleaning begins, uniformly covering the entire interior of the tank. The system ignores the actual distribution of sediment and only performs a closed-loop traversal operation. While this approach has advantages such as wide coverage and simple logical structure, it has significant limitations: in areas with concentrated sediment, the cleaning intensity may be insufficient, and the robot only briefly passes through without lingering; conversely, in areas with high cleanliness and no need for treatment, time is wasted on repeated cleaning, leading to inefficiency.

[0004] Due to factors such as uneven distribution of water flow disturbance, irregular tank structure, or local turbulence at the inlet and outlet, sediments often adhere unevenly to specific low-flow areas or corners. This heterogeneity cannot be covered by a fixed-path mechanism. If the robot cannot identify and prioritize the treatment of these areas, visible contaminant patches may remain after cleaning, reducing the quality of the secondary water supply. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent control method and system for a secondary water supply tank cleaning robot, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: an intelligent control system for a secondary water supply tank cleaning robot, comprising a sediment perception and data acquisition module, a sediment distribution modeling module, a dynamic cleaning priority determination module, a path reconstruction and scheduling module, an execution and feedback module, and a data archiving and optimization learning module; The sediment sensing and data acquisition module collects water quality data through sensors installed inside the water tank, performs preprocessing, and obtains the water tank dataset SW. The sediment distribution modeling module maps the tank dataset SW to a three-dimensional spatial grid of the tank and calculates the sediment hotspot index Hs. The dynamic cleaning priority determination module sorts the deposition hotspot index Hs of all cells, assigns a cleaning priority level Pq to each cell, and obtains a list of high-priority cleaning tasks Phigh. The path reconstruction and scheduling module, based on the high-priority cleaning task list Phigh, fully covers the initial path Lo of the cleaning robot and obtains the adjusted cleaning path trajectory vector nL. The execution and feedback module inputs the adjusted cleaning path trajectory vector nL into the robot's execution trajectory and calculates and obtains the real-time cleaning coverage Co. The data archiving and optimization learning module archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co of this cleaning, and provides feedback.

[0007] Preferably, the sediment sensing and data acquisition module includes a distributed sensing acquisition unit and a data cleaning and standardization processing unit; The distributed sensing acquisition unit has two types of sensors, namely a turbidity gradient sensor array and a sediment concentration optical probe, equidistantly arranged in three directions inside the water tank, including the vertical direction z, the horizontal direction x, and the diagonal direction y. It collects the water turbidity gradient Ct and the sediment concentration distribution index Sd, and fits them into the original dataset YW. Among them, the turbidity gradient Ct of the water body is collected by a turbidity gradient sensor array; The method for obtaining the turbidity gradient Ct of the water body is as follows: First, select points Si at equal intervals along the three-dimensional coordinate direction (x, y, z) inside the water tank, with the distance between adjacent points being Δd, to form a spatial grid; Then, the scattered light intensity per unit time is collected at each point Si and converted into a turbidity value T; finally, six adjacent points are selected around the center point Sc of the local cubic unit, and the turbidity gradient Ct of the water body is calculated. The turbidity gradient Ct of the water body is obtained by the following formula:

[0008] In the formula, T(x+Δd) represents the turbidity value collected by the right sensor in the x-axis direction, T(x-Δd) represents the turbidity value collected by the left sensor in the x-axis direction, T(y+Δd) represents the turbidity value collected by the right sensor in the y-axis direction, T(y-Δd) represents the turbidity value collected by the left sensor in the y-axis direction, T(z+Δd) represents the turbidity value collected by the right sensor in the z-axis direction, and T(z-Δd) represents the turbidity value collected by the left sensor in the z-axis direction. The sediment concentration distribution index Sd was collected using a sediment concentration optical probe. The sediment concentration distribution index Sd is obtained by setting multiple detection points at fixed intervals along the depth direction in the water tank, recording the reflected light intensity per unit area of ​​the detection points and the position of the spectral peak, and calculating the sediment concentration factor. Sediment concentration factor: First, the reflected light intensity and peak wavelength are obtained from the optical probe at each measurement point. The reflected light intensity represents the amount of light energy reflected back from the sediment layer at that location, while the peak wavelength reflects the band in the reflection spectrum where energy is concentrated, which is the characteristic response range of the sediment particles. Then, the logarithm of the ratio of reflected light intensity to peak wavelength is taken to linearize the reflection changes of different orders of magnitude, thereby reducing the error caused by the nonlinear response of the spectrum. Then, a proportionality coefficient related to the optical structure and material properties of the probe is used for correction. This proportionality coefficient is obtained through experimental calibration to ensure the comparability of measurement results under different optical paths. The resulting value is expressed as the sediment concentration factor at that measurement point. Sediment concentration factors from different detection points are stitched together to form a one-dimensional profile, which is then combined with relative depth to form a sediment concentration distribution index Sd. The sediment concentration distribution index Sd is obtained using the following formula: ; In the formula, Sd(zl) represents the sediment concentration distribution index of the zl-th layer, L represents the number of sediment concentration sampling layers, zl represents the depth sampling layer number, D(zl) represents the sediment concentration factor of the zl-th layer, and H represents the total vertical depth inside the tank. The data cleaning and standardization unit cleans and standardizes the original dataset YW to obtain the water tank dataset SW. Cleaning includes outlier removal and spatial oversampling deduplication; The outlier removal method is as follows: For the data of each sensor sampling point, calculate the turbidity difference and sediment concentration difference between the sampling point and the eight surrounding sensor nodes; compare the maximum deviation between the turbidity gradient value and sediment concentration value of the current node and the neighboring nodes to obtain the local turbidity deviation and concentration deviation. If the local turbidity deviation of a sampling point is greater than the turbidity threshold, or the concentration deviation is greater than the concentration threshold, the sampling point is marked as an outlier and removed from the dataset. The spatial oversampling deduplication method is as follows: select the coordinates of all valid sampling points within a radius r, and then aggregate the turbidity and sediment concentration data of the sampling points. All sampling points are treated as a voxel unit, and the average turbidity and average concentration of all sampling points within the voxel unit are calculated to obtain the voxelized aggregation result. By averaging the turbidity and concentration values ​​of all points, the system obtains a representative value for the local area, which is used to replace the original multi-point measurement results. The purpose of this fusion processing is to preserve the macroscopic trend characteristics of the water body and eliminate high-frequency noise caused by local disturbances, bubbles, or instantaneous fluctuations. It is equivalent to performing a physical median filter on the disturbance signal inside the water body on a spatial scale, so that subsequent analysis can better reflect the overall state of sediment distribution. The standardization process involves scaling the original dataset YW to generate a dimensionless water tank dataset SW.

[0009] The standardization formula is: ; In the formula, SWo represents the o-th data in the water tank dataset SW, YWo represents the o-th data in the original dataset YW, μYWo represents the mean of the o-th data in the original dataset YW, and σYWo represents the standard deviation of the o-th data in the original dataset YW.

[0010] Preferably, the sediment distribution modeling module includes a spatial grid division and data mapping unit and a sediment hotspot index calculation unit; The spatial grid division and data mapping unit divides the entire interior space of the water tank into n three-dimensional grid units according to a unit cube with a side length of xd. Each grid unit is numbered i∈[1,n] and has a coordinate center of (xi,yi,zi). For each data point in the water tank dataset SW, the spatial coordinates are placed into a three-dimensional grid cell to construct a local dataset within each cell; The formula is as follows: SWi={Ctk, Sdk, dk∣(xk, yk, zk)∈Vi}; In the formula, SWi represents the water tank dataset of the i-th cell grid, Vi represents the spatial volume of the i-th cell grid, (xk, yk, zk) represents the coordinates of the k-th point within the i-th cell grid, Ctk represents the water turbidity gradient at the k-th point, Sdk represents the sediment concentration distribution index at the k-th point, and dk represents the relative depth at the k-th point.

[0011] Preferably, the deposition hotspot index calculation unit extracts statistical features, including turbidity disturbance intensity pCt and concentration density pSd, from the local dataset within the cell, and calculates the deposition hotspot index Hs. The turbidity disturbance intensity pCt is obtained by: statistically analyzing the turbidity gradient values ​​Ct of all sampling points in the i-th cell grid and averaging them to obtain the turbidity disturbance intensity pCt. The concentration density pSd is obtained by: statistically analyzing the sediment concentration distribution index of all sampling points in the i-th cell grid and averaging it to obtain the concentration density pSd. The deposition hotspot index Hs is obtained using the following formula: ; In the formula, Hs(i) represents the deposition hotspot index of the i-th cell grid, pCt(i) represents the turbidity disturbance intensity in the i-th cell grid, pSd(i) represents the concentration density in the i-th cell grid, ln represents the logarithmic function, H represents the total vertical depth inside the tank, and Zdi represents the height coordinate of the center point of the i-th cell grid.

[0012] Preferably, the dynamic cleaning priority determination module includes a priority level calculation unit and a high-priority filtering and task list generation unit; The priority level calculation unit maps the deposition hotspot index Hs(i) of each cell grid i to the cleaning priority level Pq; The cleaning priority level Pq is obtained using the following formula: ; In the formula, Pq(i) represents the cleaning priority level of the i-th unit grid, n represents the total number of three-dimensional grid units, j represents the index number of all grid units in the water tank, and N represents the set priority level, for example, set to 10, which means the priority ranges from 1 to 10. The high-priority filtering and task list generation unit compares the obtained cleaning priority level Pq with the preset priority threshold Tpq, filters out all cells that satisfy Pq(i) > Tpq, and forms a high-priority cleaning task list Phigh. The screening formula is: Phigh={Gi∣Pq(i)>Tpq}; In the formula, Gi represents the i-th element mesh in the three-dimensional space of the water tank. represents a specific cubic region obtained after dividing the three-dimensional space of the water tank.

[0013] Preferably, the path reconstruction and scheduling module includes a priority area trajectory offset reconstruction unit and a path execution scheduling and trajectory compression unit; The priority area trajectory offset reconstruction unit performs spatial offset logic on the initial path Lo of the cleaning robot based on the obtained high-priority cleaning task list Phigh. The offset does not adopt a simple jump or replacement method, but is based on path interpolation adjustment to achieve dynamic offset-fusion adjustment and obtain the adjusted cleaning path trajectory vector nL. In each time node t, a trajectory adjustment item for the high-priority grid is inserted into the path instruction to form an optimized path; The adjusted cleaning path trajectory vector nL is obtained using the following formula: ; In the formula, nL(t) represents the adjusted cleaning path trajectory vector at time t, δ(Pq(i)>Tpq) represents the step function, which is 1 when Pq(i)>Tpq and 0 otherwise, and sV(i) represents the path offset vector from the current position to the i-th priority cell grid.

[0014] The path offset vector sV(i) from the current position to the i-th priority cell grid is obtained by the following formula: ; In the formula, wi represents the offset weight.

[0015] Preferably, the path execution scheduling and trajectory compression unit generates scheduling instructions and optimizes the path structure for the obtained adjusted cleaning path trajectory vector nL; its goal is to compress repetitive paths in low-priority areas, adjust the robot's turning strategy, and avoid frequent back-and-forth jumps, forming a scheduling path with strong executability and high time efficiency. The path structure optimization adopts the trajectory redundancy judgment method. For adjacent path segments with priority Pq(i) < Tpq and small spatial distance changes, a merging and compression operation is performed to reduce motion redundancy. The dispatch instruction generation includes determining when the angle θ formed by three consecutive path points A→B→C in the turning area exceeds the preset sharp turn threshold θc, and then generating a smooth turning arc segment through interpolation to reduce the speed jitter of the motion system.

[0016] Trajectory length evaluation metric: Finally, the path is evaluated, and the path length compression ratio Rl is calculated; The path length compression ratio Rl is obtained using the following formula: ; Preferably, the execution and feedback module includes a cleaning path instruction issuing unit and a real-time coverage calculation unit; The cleaning path instruction issuing unit converts the adjusted cleaning path trajectory vector nL into a sequence of robot control instructions, including displacement, steering, and spray start / stop, and drives the robot to execute the cleaning task on the corresponding path through the low-level controller; The control command structure includes: a sequence of cleaning path points; Corresponding timestamp; Printer head operating status: On / Off; Cleaning speed and steering angle settings; The real-time coverage calculation unit updates the cleaning coverage in real time during the cleaning process, compares the total area that can be cleaned in the water tank with the currently covered area, and calculates the coverage Co. ; In the formula, Cmt represents the set of areas determined to be cleaned at time t, Am represents the area of ​​the m-th cleaned area, and Atot represents the total cleaned area of ​​the inner wall of the water tank.

[0017] Preferably, the data archiving and optimization learning module packages and archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co, and stores them in the deposition learning database. Based on the sedimentation learning database, and by analyzing the archiving path and cleaning duration data, the cleaning efficiency E per unit area of ​​each cleaning grid cell is calculated. The cleaning efficiency E is obtained by the ratio of the cleaning area of ​​the grid cell to the total time the robot spends in the grid cell; The obtained cleaning efficiency E is compared with the preset efficiency threshold Te to determine the cleaning status of the grid cells; When the cleaning efficiency E > the efficiency threshold Te, it indicates that the cleaning status of the grid cell is normal. When the cleaning efficiency E ≤ efficiency threshold Te, it indicates that the cleaning status of the grid cell is abnormal and is marked as a modification area; and the cleaning strategy is executed: increase the number of times the cleaning path is repeatedly covered, increase the nozzle pressure setting value in this area, and extend the robot's dwell time.

[0018] A method for intelligent control of a secondary water supply tank cleaning robot includes the following steps: Step 1: The sediment sensing and data acquisition module collects water quality data through sensors installed inside the water tank and preprocesses it to obtain the water tank dataset SW. Step 2: The sediment distribution modeling module maps the tank dataset SW to a 3D spatial grid of the tank and calculates the sediment hotspot index Hs. Step 3: The dynamic cleaning priority determination module sorts the deposition hotspot index Hs of all cells and assigns a cleaning priority level Pq to each cell, and obtains the high-priority cleaning task list Phigh. Step 4: The path reconstruction and scheduling module, based on the high-priority cleaning task list Phigh, fully covers the initial path Lo of the cleaning robot to obtain the adjusted cleaning path trajectory vector nL. Step 5: The execution and feedback module inputs the adjusted cleaning path trajectory vector nL into the robot's execution trajectory and calculates and obtains the real-time cleaning coverage Co. Step Six: The data archiving and optimization learning module archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co of this cleaning process, and provides feedback.

[0019] This invention provides an intelligent control method and system for a secondary water supply tank cleaning robot, which has the following beneficial effects: (1) During system operation, the system spatially models the water state inside the tank through the sediment perception and data acquisition module. The system can identify specific areas where sediments accumulate, instead of relying on the traditional blind traversal mode of fixed paths, thus shifting the cleaning behavior from a comprehensive approach to a focused approach. The system introduces dynamic judgment logic of sediment hotspot index and cleaning priority level, which can actively identify areas inside the tank that are difficult to clean and have dense sediments, and prioritize these areas in the robot path for processing, thereby improving the targeting and coverage of the cleaning action.

[0020] Through the data archiving and optimization learning module, the system stores and learns the deposition distribution characteristics, trajectory execution records and coverage data formed in each round of cleaning. It extracts inefficient areas and task performance bottlenecks from these data and adjusts the pressure strategy and path structure in advance in subsequent cleaning cycles to gradually optimize operating efficiency and cleaning effect.

[0021] (2) By dividing the internal space of the water tank into three-dimensional cubic grid cells with fixed side lengths, the sedimentation state is no longer expressed by fuzzy or overall average methods, but by achieving fixed-dimensional modeling in the spatial dimension. Each cell has an independent number and coordinate center, which facilitates precise positioning and focused analysis. This method improves the spatial representation of sediment distribution inside the water tank and provides a structured reference at the regional distribution level for the formulation of cleaning paths. A local dataset was constructed within each cell and statistically processed to extract two key indicators: "turbidity disturbance intensity" and "concentration density". This processing method takes into account both the characteristics of water flow disturbance and sediment accumulation behavior, thus being able to truly reflect the sedimentation formation mechanism and cleaning difficulty of each area, solving the problem that traditional methods cannot identify the local cleaning focus.

[0022] (3) By establishing a mapping relationship between each cleaning action and the total cleanable area on the inner wall of the water tank, the current cleaned area is compared in real time, and a continuous coverage feedback curve is constructed. This enables the system to dynamically adjust the task flow based on the coverage during the cleaning process. It can automatically terminate the task after the coverage reaches the target standard, and can also identify the missed areas for backtracking coverage, ensuring that the path execution results have an objective quantitative evaluation basis.

[0023] The data archiving and optimization learning module establishes a structured sedimentation learning database based on historical records of sedimentation hotspot index Hs, trajectory vector nL, and coverage Co. It further identifies "difficult-to-clean areas" and "efficiency anomaly units" by modeling and evaluating the cleaning efficiency of unit grids.

[0024] (4) By deploying a multi-type distributed sensor array inside the water tank, the system can simultaneously sense the turbidity gradient and sediment concentration distribution, and construct a high-resolution three-dimensional spatial dataset to achieve a comprehensive understanding of sedimentation trends and hotspot areas. This mechanism overcomes the problems of traditional cleaning schemes being "invisible" and "difficult to judge" the sediment distribution, and lays a dynamic data foundation for the precise scheduling of subsequent cleaning paths.

[0025] By quantifying the depositional hotspot index Hs, the spatial depositional density and turbidity disturbance state are integrated for calculation, thereby constructing a priority determination system Pq to effectively identify high-depositional areas requiring priority treatment. Combined with the initial path Lo, the system can introduce targeted trajectory offsets and path interpolation based on the original path, achieving dynamic reconstruction of the cleaning path. This transforms cleaning from a purely mechanical, full-coverage process into proactive coverage and resource allocation to "risk areas," significantly improving cleaning efficiency and the targeted nature of depositional treatment. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the intelligent control system for a secondary water supply tank cleaning robot according to the present invention. Figure 2 This is a schematic diagram of the steps of an intelligent control method for a secondary water supply tank cleaning robot according to the present invention; Figure 3 This is a schematic diagram of the process for obtaining the cleaning efficiency of the present invention; Figure 4 This is a trend chart of the deposition hotspot index of the present invention. Detailed Implementation

[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] Example 1 This invention provides an intelligent control system for a secondary water supply tank cleaning robot. Please refer to [link / reference]. Figures 1 to 4It includes a sediment sensing and data acquisition module, a sediment distribution modeling module, a dynamic cleaning priority determination module, a path reconstruction and scheduling module, an execution and feedback module, and a data archiving and optimization learning module; The sediment sensing and data acquisition module collects water quality data through sensors installed inside the water tank, performs preprocessing, and obtains the water tank dataset SW. The sediment distribution modeling module maps the tank dataset SW to a three-dimensional spatial grid of the tank and calculates the sediment hotspot index Hs. The dynamic cleaning priority determination module sorts the deposition hotspot index Hs of all cells, assigns a cleaning priority level Pq to each cell, and obtains a list of high-priority cleaning tasks Phigh. The path reconstruction and scheduling module, based on the high-priority cleaning task list Phigh, fully covers the initial path Lo of the cleaning robot and obtains the adjusted cleaning path trajectory vector nL. The execution and feedback module inputs the adjusted cleaning path trajectory vector nL into the robot's execution trajectory and calculates and obtains the real-time cleaning coverage Co. The data archiving and optimization learning module archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co of this cleaning, and provides feedback.

[0029] In this embodiment, the system spatially models the water state inside the tank using a sediment sensing and data acquisition module. This allows the system to identify specific areas where sediment accumulates, moving away from the traditional blind traversal of fixed paths and shifting the cleaning process from a comprehensive approach to a focused one. A dynamic judgment logic based on sediment hotspot indices and cleaning priority levels is introduced, enabling the system to proactively identify areas inside the tank that are difficult to clean and have dense sediment, prioritizing these areas for processing within the robot's path. This enhances the targeting and effectiveness of the cleaning operation.

[0030] Through the data archiving and optimization learning module, the system stores and learns the deposition distribution characteristics, trajectory execution records and coverage data formed in each round of cleaning. It extracts inefficient areas and task performance bottlenecks from these data and adjusts the pressure strategy and path structure in advance in subsequent cleaning cycles to gradually optimize operating efficiency and cleaning effect.

[0031] This invention enhances the adaptability of the water tank cleaning system to actual working conditions. The modules designed in this invention possess the ability to identify and adapt to various factors such as non-uniform sediment distribution, complex water tank structure, and differences in hydraulic disturbances. System control no longer relies on a single strategy but is optimized in real time based on the on-site conditions, making it suitable for water quality assurance needs in various scenarios.

[0032] Example 2 This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 and Figure 3 Specifically: the sediment sensing and data acquisition module includes a distributed sensing acquisition unit and a data cleaning and standardization processing unit; The distributed sensing acquisition unit has two types of sensors, namely a turbidity gradient sensor array and a sediment concentration optical probe, equidistantly arranged in three directions inside the water tank, including the vertical direction z, the horizontal direction x, and the diagonal direction y. It collects the water turbidity gradient Ct and the sediment concentration distribution index Sd, and fits them into the original dataset YW. Among them, the turbidity gradient Ct of the water body is collected by a turbidity gradient sensor array; The sediment concentration distribution index Sd was collected using a sediment concentration optical probe. The data cleaning and standardization unit cleans and standardizes the original dataset YW to obtain the water tank dataset SW. Cleaning includes outlier removal and spatial oversampling deduplication; The outlier removal method is as follows: For the data of each sensor sampling point, calculate the turbidity difference and sediment concentration difference between the sampling point and the eight surrounding sensor nodes; compare the maximum deviation between the turbidity gradient value and sediment concentration value of the current node and the neighboring nodes to obtain the local turbidity deviation and concentration deviation. If the local turbidity deviation of a sampling point is greater than the turbidity threshold, or the concentration deviation is greater than the concentration threshold, the sampling point is marked as an outlier and removed from the dataset. The spatial oversampling deduplication method is as follows: select the coordinates of all valid sampling points within a radius r, and then aggregate the turbidity and sediment concentration data of the sampling points. All sampling points are treated as a voxel unit, and the average turbidity and average concentration of all sampling points within the voxel unit are calculated to obtain the voxelized aggregation result. By averaging the turbidity and concentration values ​​of all points, a representative value of the local area is obtained to replace the original multi-point measurement results. The standardization process involves scaling the original dataset YW to generate a dimensionless water tank dataset SW.

[0033] In this embodiment, by equidistantly deploying turbidity gradient sensors and sediment concentration optical probes along the three-dimensional (x, y, z) directions of the water tank, the system can comprehensively perceive the sedimentary characteristics of different layers and directions within the water body. This eliminates reliance on limited locations or single-plane sampling, enhancing the perception of corner areas, stagnant water zones, and settling areas, thus providing a complete foundation for subsequent accurate modeling. By comparing the turbidity and concentration deviations between each sampling point and its eight adjacent points, local abrupt changes are identified and eliminated. This outlier elimination strategy can offset the impact of non-realistic physical phenomena such as bubble interference, light path refraction, and transient disturbances on data accuracy, thereby significantly improving the stability and reliability of sedimentary feature identification.

[0034] To address the potential issues of data duplication and oversampling arising from high-density sampling points, the system delineates sampling points into voxel units based on spatial radius, performs fusion calculations, and outputs local average turbidity and concentration values. This approach preserves spatial distribution characteristics while reducing computational redundancy. This processing makes the data representation more representative and interpretable, while also providing higher-quality input for the mesh modeling stage.

[0035] The original data differs in both spatial and numerical scales. To ensure comparability of data from different measurement points, this embodiment performs normalization and standardization on the processed data, outputting a water tank dataset in a unified format. This process not only eliminates the interference of unit scale differences on the analysis model but also establishes a unified data foundation for subsequent calculations of spatial indicators such as the sedimentary hotspot index.

[0036] Example 3 This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 and Figure 4 Specifically: the sediment distribution modeling module includes a spatial grid division and data mapping unit and a sediment hotspot index calculation unit; The spatial grid division and data mapping unit divides the entire interior space of the water tank into n three-dimensional grid units according to a unit cube with a side length of xd. Each grid unit is numbered i∈[1,n] and has a coordinate center of (xi,yi,zi). For each data point in the water tank dataset SW, the spatial coordinates are placed into a three-dimensional grid cell to construct a local dataset within each cell; The formula is as follows: SWi={Ctk, Sdk, dk∣(xk, yk, zk)∈Vi}; In the formula, SWi represents the water tank dataset of the i-th cell grid, Vi represents the spatial volume of the i-th cell grid, (xk, yk, zk) represents the coordinates of the k-th point within the i-th cell grid, Ctk represents the water turbidity gradient at the k-th point, Sdk represents the sediment concentration distribution index at the k-th point, and dk represents the relative depth at the k-th point.

[0037] The deposition hotspot index calculation unit extracts statistical features from the local dataset within the cell, including turbidity disturbance intensity pCt and concentration density pSd, and calculates the deposition hotspot index Hs. The turbidity disturbance intensity pCt is obtained by: statistically analyzing the turbidity gradient values ​​Ct of all sampling points in the i-th cell grid and averaging them to obtain the turbidity disturbance intensity pCt. The concentration density pSd is obtained by: statistically analyzing the sediment concentration distribution index of all sampling points in the i-th cell grid and averaging it to obtain the concentration density pSd. The deposition hotspot index Hs is obtained using the following formula: ; In the formula, Hs(i) represents the deposition hotspot index of the i-th cell grid, pCt(i) represents the turbidity disturbance intensity in the i-th cell grid, pSd(i) represents the concentration density in the i-th cell grid, ln represents the logarithmic function, H represents the total vertical depth inside the tank, and Zdi represents the height coordinate of the center point of the i-th cell grid.

[0038] In this embodiment, the internal space of the water tank is divided into three-dimensional cubic grid cells with fixed side lengths. Instead of using fuzzy or overall average methods to represent the sedimentation state, fixed-frame modeling in the spatial dimension is achieved. Each cell has an independent number and coordinate center, facilitating precise positioning and focused analysis. This approach enhances the spatial representation of sediment distribution within the water tank, providing a structured reference at the regional distribution level for determining cleaning paths.

[0039] In this embodiment, a local dataset was constructed within each unit and statistically processed to extract two key indicators: "turbidity disturbance intensity" and "concentration density." This processing method takes into account both water flow disturbance characteristics and sediment accumulation behavior, thereby accurately reflecting the sediment formation mechanism and cleaning difficulty of each area, and solving the problem that traditional methods cannot identify the focus of local cleaning.

[0040] In this embodiment, the vertical height coordinate of the water tank is introduced as a depth adjustment factor in the calculation of the sedimentation hotspot index. Combined with turbidity disturbance and concentration level, a logarithmic function is used to express the spatial trend of sediment accumulation, fully reflecting the physical relationship between sedimentation behavior and the gravity field of the water body. This spatially coupled index system not only has stronger interpretability but also effectively distinguishes the differences in sedimentation risk at different height levels, thus providing a scientific basis for path optimization and promoting the transformation of cleaning behavior from "equal coverage" to "on-demand scheduling".

[0041] Example 4 This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 3 Specifically: the dynamic cleaning priority determination module includes a priority level calculation unit and a high-priority filtering and task list generation unit; The priority level calculation unit maps the deposition hotspot index Hs(i) of each cell grid i to the cleaning priority level Pq; The cleaning priority level Pq is obtained using the following formula: ; In the formula, Pq(i) represents the cleaning priority level of the i-th unit grid, n represents the total number of three-dimensional grid units, j represents the index number of all grid units in the water tank, and N represents the number of priority levels set. The high-priority filtering and task list generation unit compares the obtained cleaning priority level Pq with the preset priority threshold Tpq, filters out all cells that satisfy Pq(i) > Tpq, and forms a high-priority cleaning task list Phigh. The screening formula is: Phigh={Gi∣Pq(i)>Tpq}; In the formula, Gi represents the i-th unit grid in the three-dimensional space of the water tank.

[0042] The path reconstruction and scheduling module includes a priority area trajectory offset reconstruction unit and a path execution scheduling and trajectory compression unit; The priority area trajectory offset reconstruction unit performs spatial offset logic on the initial path Lo of the cleaning robot based on the obtained high-priority cleaning task list Phigh, and obtains the adjusted cleaning path trajectory vector nL. In each time node t, a trajectory adjustment item for the high-priority grid is inserted into the path instruction to form an optimized path; The adjusted cleaning path trajectory vector nL is obtained using the following formula: ; In the formula, nL(t) represents the adjusted cleaning path trajectory vector at time t, δ(Pq(i)>Tpq) represents the step function, which is 1 when Pq(i)>Tpq and 0 otherwise, and sV(i) represents the path offset vector from the current position to the i-th priority cell grid.

[0043] The path execution scheduling and trajectory compression unit generates scheduling instructions and optimizes the path structure based on the acquired adjusted cleaning path trajectory vector nL; The path structure optimization adopts the trajectory redundancy judgment method. For adjacent path segments with priority Pq(i) < Tpq, a merging and compression operation is performed to reduce motion redundancy. The dispatch instruction generation includes determining when the angle θ formed by three consecutive path points A→B→C in the turning area exceeds the preset sharp turn threshold θc, and then generating a smooth turning arc segment through interpolation to reduce the speed jitter of the motion system.

[0044] In this embodiment, instead of employing the traditional equal-weight coverage strategy, a cleaning priority level system is constructed using the deposition hotspot index as the core parameter, giving each spatial grid cell a differentiated cleaning response level. This mechanism effectively identifies key deposition areas inside the tank and assigns different levels of cleaning attention through mapping rules, constructing a responsive cleaning strategy oriented towards deposition distribution characteristics, thus overcoming the limitations of the traditional "average" processing of cleaning paths.

[0045] At the path planning level, the system automatically performs spatial offset and target injection on the original path Lo of the cleaning robot based on the selected high-priority areas. Instead of mechanically repeating fixed trajectories, the system updates the trajectory according to the execution status at each time point, proactively inserting heavily deposited areas into the motion path. This path interpolation and offset strategy ensures that the cleaning path has the ability to proactively cover "hotspot areas," transforming path execution from static scheduling to dynamic response, thus adapting to the spatial heterogeneity of the water tank's deposition state.

[0046] This embodiment further compresses and optimizes the path based on the obtained adjusted trajectory vector nL. By identifying and merging low-priority redundant path segments, the repetitive movements of the robot in non-critical areas are reduced. At the same time, an interpolation smoothing mechanism is introduced in the turning segments of the path to effectively avoid system vibration and trajectory deviation caused by sharp turns, thereby improving the stability and energy efficiency of the robot's execution and enhancing the system's cleaning performance under complex paths.

[0047] Example 5 This embodiment is an explanation based on Embodiment 4. Please refer to it. Figure 1 Specifically: the execution and feedback module includes a cleaning path instruction issuance unit and a real-time coverage calculation unit; The cleaning path instruction issuing unit converts the adjusted cleaning path trajectory vector nL into a sequence of robot control instructions, including displacement, steering, and spray start / stop, and drives the robot to execute the cleaning task on the corresponding path through the low-level controller; The control command structure includes: a sequence of cleaning path points; Corresponding timestamp; Printer head operating status: On / Off; Cleaning speed and steering angle settings; The real-time coverage calculation unit updates the cleaning coverage in real time during the cleaning process, compares the total area that can be cleaned in the water tank with the currently covered area, and calculates the coverage Co. ; In the formula, Cmt represents the set of areas determined to be cleaned at time t, Am represents the area of ​​the m-th cleaned area, and Atot represents the total cleaned area of ​​the inner wall of the water tank.

[0048] The data archiving and optimization learning module packages and archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co, and stores them in the deposition learning database. Based on the sedimentation learning database, and by analyzing the archiving path and cleaning duration data, the cleaning efficiency E per unit area of ​​each cleaning grid cell is calculated. The cleaning efficiency E is obtained by the ratio of the cleaning area of ​​the grid cell to the total time the robot spends in the grid cell; The obtained cleaning efficiency E is compared with the preset efficiency threshold Te to determine the cleaning status of the grid cells; When the cleaning efficiency E > the efficiency threshold Te, it indicates that the cleaning status of the grid cell is normal. When the cleaning efficiency E ≤ efficiency threshold Te, it indicates that the cleaning status of the grid cell is abnormal and is marked as a modification area; and the cleaning strategy is executed: increase the number of times the cleaning path is repeatedly covered, increase the area nozzle pressure setting value, and extend the robot's dwell time.

[0049] In this embodiment, the adjusted cleaning path trajectory vector nL is transformed into a robot control command sequence, specifying control parameters including displacement, steering, spray start / stop, speed, and turning angle. A complete command chain is generated through an integrated cleaning path command issuing unit, ensuring the accuracy, continuity, and controllability of the cleaning path execution. This structure overcomes the limitations of previous coarse-grained cleaning path control, achieving closed-loop integration between fine-grained path management and dynamic adaptive scheduling. It can handle real-world situations such as frequent turning points in complex spaces and uneven target grid distribution.

[0050] This embodiment establishes a mapping relationship between each cleaning action and the total washable area on the inner wall of the water tank, compares the current cleaned area in real time, and constructs a continuous coverage feedback curve. This allows the system to dynamically adjust the task flow based on the coverage during the cleaning process. It can automatically terminate the task after the coverage reaches the target standard, and can also identify missed areas for backtracking coverage, ensuring that the path execution results have an objective and quantitative evaluation basis.

[0051] The data archiving and optimization learning module establishes a structured sedimentation learning database based on historical records of sedimentation hotspot index Hs, trajectory vector nL, and coverage Co. It further identifies "difficult-to-clean areas" and "efficiency anomaly units" by modeling and evaluating the cleaning efficiency of unit grids.

[0052] Example 6 A method for intelligent control of a secondary water supply tank cleaning robot, please refer to... Figure 2 Specifically, it includes the following steps: Step 1: The sediment sensing and data acquisition module collects water quality data through sensors installed inside the water tank and preprocesses it to obtain the water tank dataset SW. Step 2: The sediment distribution modeling module maps the tank dataset SW to a 3D spatial grid of the tank and calculates the sediment hotspot index Hs. Step 3: The dynamic cleaning priority determination module sorts the deposition hotspot index Hs of all cells and assigns a cleaning priority level Pq to each cell, and obtains the high-priority cleaning task list Phigh. Step 4: The path reconstruction and scheduling module, based on the high-priority cleaning task list Phigh, fully covers the initial path Lo of the cleaning robot to obtain the adjusted cleaning path trajectory vector nL. Step 5: The execution and feedback module inputs the adjusted cleaning path trajectory vector nL into the robot's execution trajectory and calculates and obtains the real-time cleaning coverage Co. Step Six: The data archiving and optimization learning module archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co of this cleaning process, and provides feedback.

[0053] In this embodiment, by deploying a multi-type distributed sensor array inside the water tank, the system can simultaneously sense the turbidity gradient and sediment concentration distribution, and construct a high-resolution three-dimensional spatial dataset, achieving a comprehensive understanding of sedimentation trends and hotspot areas. This mechanism overcomes the problems of traditional cleaning schemes being "invisible" and "difficult to judge" regarding sediment distribution, laying a dynamic data foundation for the precise scheduling of subsequent cleaning paths.

[0054] By quantifying the depositional hotspot index Hs, the spatial depositional density and turbidity disturbance state are integrated for calculation, thereby constructing a priority determination system Pq to effectively identify high-depositional areas requiring priority treatment. Combined with the initial path Lo, the system can introduce targeted trajectory offsets and path interpolation based on the original path, achieving dynamic reconstruction of the cleaning path. This transforms cleaning from a purely mechanical, full-coverage process into proactive coverage and resource allocation to "risk areas," significantly improving cleaning efficiency and the targeted nature of depositional treatment.

[0055] By collecting real-time data on the cleaning execution coverage status, a "cleaning coverage Co" is constructed as a dynamic feedback indicator. Based on this, the system can determine the degree of task completion and missed areas, enabling phased termination of tasks and retrospective supplementation of missed areas. Simultaneously, deposition hotspot indices, path trajectories, and coverage data are archived to establish a deposition learning database. By assessing cleaning efficiency and identifying "difficult-to-clean areas," the system can automatically match higher pressure parameters, longer dwell times, or repeated coverage paths in the next task, possessing self-learning and continuous optimization capabilities. This allows for an intelligent leap from single-task execution to long-term strategy evolution.

[0056] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent control system for a secondary water supply tank cleaning robot, characterized in that: It includes a sediment sensing and data acquisition module, a sediment distribution modeling module, a dynamic cleaning priority determination module, a path reconstruction and scheduling module, an execution and feedback module, and a data archiving and optimization learning module; The sediment sensing and data acquisition module collects water quality data through sensors installed inside the water tank, performs preprocessing, and obtains the water tank dataset. The sediment distribution modeling module maps the tank dataset to a three-dimensional spatial grid of the tank and calculates the sediment hotspot index. The sediment distribution modeling module includes a spatial grid division and data mapping unit and a sediment hotspot index calculation unit. The deposition hotspot index calculation unit extracts statistical features from the local dataset within the cell, including turbidity disturbance intensity pCt and concentration density pSd, and calculates the deposition hotspot index Hs. The turbidity disturbance intensity pCt is obtained by: statistically analyzing the turbidity gradient values ​​Ct of all sampling points in the i-th cell grid and averaging them to obtain the turbidity disturbance intensity pCt. The concentration density pSd is obtained by: statistically analyzing the sediment concentration distribution index of all sampling points in the i-th cell grid and averaging it to obtain the concentration density pSd. The deposition hotspot index Hs is obtained using the following formula: ； In the formula, Hs(i) represents the deposition hotspot index of the i-th cell grid, pCt(i) represents the turbidity disturbance intensity in the i-th cell grid, pSd(i) represents the concentration density in the i-th cell grid, ln represents the logarithmic function, H represents the total vertical depth inside the tank, and Zdi represents the height coordinate of the center point of the i-th cell grid. The dynamic cleaning priority determination module sorts the deposition hotspot indices of all three-dimensional spatial grids and assigns cleaning priority levels to obtain a list of high-priority cleaning tasks. The path reconstruction and scheduling module, based on a high-priority cleaning task list, fully covers the initial path of the cleaning robot and obtains the adjusted cleaning path trajectory vector. The path reconstruction and scheduling module includes a priority area trajectory offset reconstruction unit and a path execution scheduling and trajectory compression unit; The priority area trajectory offset reconstruction unit performs spatial offset logic on the initial path Lo of the cleaning robot based on the obtained high-priority cleaning task list Phigh, and obtains the adjusted cleaning path trajectory vector nL. In each time node t, a trajectory adjustment item for the high-priority grid is inserted into the path instruction to form an optimized path; The adjusted cleaning path trajectory vector nL is obtained using the following formula: ； In the formula, nL(t) represents the adjusted cleaning path trajectory vector at time t, δ(Pq(i)>Tpq) represents the step function, Pq(i) represents the cleaning priority level of the i-th cell grid, Tpq represents the preset priority threshold, which is 1 when Pq(i)>Tpq and 0 otherwise, and sV(i) represents the path offset vector from the current position to the i-th priority cell grid. The execution and feedback module inputs the adjusted cleaning path trajectory vector into the robot's execution trajectory to calculate and obtain the real-time cleaning coverage. The data archiving and optimization learning module archives the deposition hotspot index, the adjusted cleaning path trajectory vector, and the real-time cleaning coverage of this cleaning process, and provides feedback.

2. The intelligent control system for a secondary water supply tank cleaning robot according to claim 1, characterized in that: The sediment sensing and data acquisition module includes a distributed sensing acquisition unit and a data cleaning and standardization processing unit; The distributed sensing acquisition unit has two types of sensors, namely a turbidity gradient sensor array and a sediment concentration optical probe, equidistantly arranged in three directions inside the water tank, including the vertical direction z, the horizontal direction x, and the diagonal direction y. It collects the water turbidity gradient Ct and the sediment concentration distribution index Sd, and fits them into the original dataset YW. Among them, the turbidity gradient Ct of the water body is collected by a turbidity gradient sensor array; The sediment concentration distribution index Sd was collected using a sediment concentration optical probe. The data cleaning and standardization unit cleans and standardizes the original dataset YW to obtain the water tank dataset SW. Cleaning includes outlier removal and spatial oversampling deduplication; The outlier removal method is as follows: For the data of each sensor sampling point, calculate the turbidity difference and sediment concentration difference between the sampling point and the eight surrounding sensor nodes; compare the maximum deviation between the turbidity gradient value and sediment concentration value of the current node and the neighboring nodes to obtain the local turbidity deviation and concentration deviation. If the local turbidity deviation of a sampling point is greater than the turbidity threshold, or the concentration deviation is greater than the concentration threshold, the sampling point is marked as an outlier and removed from the dataset. The spatial oversampling deduplication method is as follows: select the coordinates of all valid sampling points within a radius r, and then aggregate the turbidity and sediment concentration data of the sampling points. All sampling points are treated as a voxel unit, and the average turbidity and average concentration of all sampling points within the voxel unit are calculated to obtain the voxelized aggregation result. By averaging the turbidity and concentration values ​​of all points, a representative value of the local area is obtained to replace the original multi-point measurement results. The standardization process involves scaling the original dataset YW to generate a dimensionless water tank dataset SW.

3. The intelligent control system for a secondary water supply tank cleaning robot according to claim 2, characterized in that: The spatial grid division and data mapping unit divides the entire interior space of the water tank into n three-dimensional grid units according to a unit cube with a side length of xd. Each grid unit is numbered i∈[1,n] and has a coordinate center of (xi,yi,zi). For each data point in the water tank dataset SW, the spatial coordinates are placed into a three-dimensional grid cell to construct a local dataset within each cell; The formula is as follows: SWi={Ctk, Sdk, dk∣(xk, yk, zk)∈Vi}; In the formula, SWi represents the water tank dataset of the i-th cell grid, Vi represents the spatial volume of the i-th cell grid, (xk, yk, zk) represents the coordinates of the k-th point within the i-th cell grid, Ctk represents the water turbidity gradient at the k-th point, Sdk represents the sediment concentration distribution index at the k-th point, and dk represents the relative depth at the k-th point.

4. The intelligent control system for a secondary water supply tank cleaning robot according to claim 3, characterized in that: The dynamic cleaning priority determination module includes a priority level calculation unit and a high-priority filtering and task list generation unit; The priority level calculation unit maps the deposition hotspot index Hs(i) of each cell grid i to the cleaning priority level Pq; The cleaning priority level Pq is obtained using the following formula: ； In the formula, Pq(i) represents the cleaning priority level of the i-th unit grid, n represents the total number of three-dimensional grid units, j represents the index number of all grid units in the water tank, and N represents the number of priority levels set. The high-priority filtering and task list generation unit compares the obtained cleaning priority level Pq with the preset priority threshold Tpq, filters out all cells that satisfy Pq(i) > Tpq, and forms a high-priority cleaning task list Phigh. The screening formula is: Phigh={Gi∣Pq(i)>Tpq}; In the formula, Gi represents the i-th unit grid in the three-dimensional space of the water tank.

5. The intelligent control system for a secondary water supply tank cleaning robot according to claim 4, characterized in that: The path execution scheduling and trajectory compression unit generates scheduling instructions and optimizes the path structure based on the acquired adjusted cleaning path trajectory vector nL; The path structure optimization adopts the trajectory redundancy judgment method. For adjacent path segments with priority Pq(i) < Tpq, a merging and compression operation is performed to reduce motion redundancy. The dispatch instruction generation includes determining when the angle θ formed by three consecutive path points A→B→C in the turning area exceeds the preset sharp turn threshold θc, and then generating a smooth turning arc segment through interpolation to reduce the speed jitter of the motion system.

6. The intelligent control system for a secondary water supply tank cleaning robot according to claim 5, characterized in that: The execution and feedback module includes a cleaning path instruction issuance unit and a real-time coverage calculation unit; The cleaning path instruction issuing unit converts the adjusted cleaning path trajectory vector nL into a sequence of robot control instructions, including displacement, steering, and spray start / stop, and drives the robot to perform the cleaning task on the corresponding path through the controller; The control command structure includes: a sequence of cleaning path points; Corresponding timestamp; Printer head operating status: On / Off; Cleaning speed and steering angle settings; The real-time coverage calculation unit updates the cleaning coverage in real time during the cleaning process, compares the total area that can be cleaned in the water tank with the currently covered area, and calculates the coverage Co. ； In the formula, Ctm represents the set of areas determined to be cleaned at time t, Am represents the area of ​​the m-th cleaned area, and Atot represents the total cleaned area of ​​the inner wall of the water tank.

7. The intelligent control system for a secondary water supply tank cleaning robot according to claim 6, characterized in that: The data archiving and optimization learning module packages and archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co, and stores them in the deposition learning database. Based on the sedimentation learning database, and by analyzing the archiving path and cleaning duration data, the cleaning efficiency E per unit area of ​​each cleaning grid cell is calculated. The cleaning efficiency E is obtained by the ratio of the cleaning area of ​​the grid cell to the total time the robot spends in the grid cell; The obtained cleaning efficiency E is compared with the preset efficiency threshold Te to determine the cleaning status of the grid cells; When the cleaning efficiency E > the efficiency threshold Te, it indicates that the cleaning status of the grid cell is normal. When the cleaning efficiency E ≤ efficiency threshold Te, it indicates that the cleaning status of the grid cell is abnormal and is marked as a modification area; and the cleaning strategy is executed: increase the number of times the cleaning path is repeatedly covered, increase the area nozzle pressure setting value, and extend the robot's dwell time.

8. A method for intelligent control of a secondary water supply tank cleaning robot, applied to the intelligent control system of a secondary water supply tank cleaning robot as described in any one of claims 1 to 7, characterized in that: Includes the following steps: Step 1: The sediment sensing and data acquisition module collects water quality data through sensors installed inside the water tank and preprocesses it to obtain the water tank dataset SW. Step 2: The sediment distribution modeling module maps the tank dataset SW to a 3D spatial grid of the tank and calculates the sediment hotspot index Hs. Step 3: The dynamic cleaning priority determination module sorts the deposition hotspot index Hs of all cells and assigns a cleaning priority level Pq to each cell, and obtains the high-priority cleaning task list Phigh. Step 4: The path reconstruction and scheduling module, based on the high-priority cleaning task list Phigh, fully covers the initial path Lo of the cleaning robot to obtain the adjusted cleaning path trajectory vector nL. Step 5: The execution and feedback module inputs the adjusted cleaning path trajectory vector nL into the robot's execution trajectory and calculates and obtains the real-time cleaning coverage Co. Step Six: The data archiving and optimization learning module archives the deposition hotspot index Hs, the adjusted cleaning path trajectory vector nL, and the real-time cleaning coverage Co of this cleaning process, and provides feedback.

Images