Control method and device of window cleaning robot, window cleaning robot, medium and product

Abstract

The application provides a window cleaning robot control method and device, a window cleaning robot, a medium and a product, and relates to the technical field of robots. The method comprises the following steps: obtaining surface characteristics of a surface to be cleaned to which a window cleaning robot is adsorbed, wherein the surface characteristics comprise surface width information, boundary type information and boundary combination state; selecting a corresponding path planning strategy according to the surface characteristics; and controlling the window cleaning robot to perform a cleaning operation on the surface to be cleaned based on the selected path planning strategy, so that the window cleaning robot can adapt to different surfaces to be cleaned, thereby effectively avoiding problems such as path mode mismatch, repeated wiping, local omission or efficiency reduction, and improving the completeness of cleaning coverage and operation efficiency.

Description

Technical Field

[0001] This application relates to the field of robotics technology, and in particular to a control method, device, window cleaning robot, medium, and product for a window cleaning robot. Background Technology

[0002] Window cleaning robots, as intelligent home cleaning devices, have been widely used for automated cleaning of high-rise building glass and home windows. With the development of sensor technology and path planning algorithms, existing window cleaning robots can typically detect window edges using collision sensors, optical sensors, or posture sensors, and achieve round-trip full-coverage cleaning in combination with preset path patterns. Common path patterns include zigzag, N-shaped, and horizontal or vertical round-trip paths.

[0003] In existing technologies, window cleaning robots, after adhering to the window surface, use sensors to detect edges and identify framed or frameless boundaries. They then clean back and forth along a preset zigzag or N-shaped path. When encountering edges, they can perform operations such as avoidance, turning, or stopping. After cleaning is completed, they perform edge cleaning and then return to the starting point.

[0004] However, the above cleaning methods only distinguish between framed and frameless windows. When dealing with narrow windows, narrow windows with frames on both sides, or windows with inconsistent boundary shapes, problems such as mismatched path patterns, repeated wiping, partial omissions, or decreased efficiency may easily occur. Summary of the Invention

[0005] This application provides a control method, device, robot, medium, and product for a window cleaning robot. By acquiring surface features including surface width information, boundary type information, and boundary combination state, and selecting a matching path planning strategy accordingly, the window cleaning robot can adapt to different surfaces to be cleaned, thereby effectively avoiding problems such as path pattern mismatch, repeated wiping, partial omissions, or decreased efficiency, and improving the integrity of cleaning coverage and operational efficiency.

[0006] In a first aspect, this application provides a control method for a window cleaning robot, the method comprising:

[0007] The surface features of the surface to be cleaned by the window cleaning robot are obtained, including surface width information, boundary type information, and boundary combination state.

[0008] The corresponding path planning strategy is selected based on the surface characteristics, and the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

[0009] In contrast to existing technologies that only distinguish between "framed windows" and "frameless windows," this approach can easily lead to path pattern mismatches, repeated wiping, partial omissions, or decreased efficiency when dealing with narrow windows, narrow windows with frames on both sides, or windows with inconsistent boundary shapes. This application introduces a more refined surface feature recognition mechanism, recognizing not only boundary type information (framed / frameless) but also surface width information and boundary combination states. By recognizing surface width information, the window cleaning robot can distinguish between different types of surfaces to be cleaned, such as wide windows and narrow windows, thus identifying window types that existing technologies cannot distinguish, such as narrow windows and narrow windows with frames on both sides. Furthermore, by recognizing boundary combination states, the window cleaning robot can identify surfaces with inconsistent boundary shapes. Therefore, through these two added dimensions, the window cleaning robot can automatically adapt to diverse surfaces to be cleaned, significantly expanding its applicable scenarios and improving its cleaning adaptability to complex surfaces.

[0010] Furthermore, this application dynamically selects the corresponding path planning strategy based on the acquired surface features, rather than executing a fixed template path. This decision-making method ensures a precise match between the path pattern and the surface features to be cleaned, effectively eliminating edge omissions and uncovered areas caused by mismatches between the path pattern and the type of surface to be cleaned, thus improving the integrity of cleaning coverage. It also avoids ineffective turns and repeated wiping caused by fixed template paths in narrow window or asymmetrical boundary scenarios, shortening the overall cleaning time. The execution of this differentiated path strategy enables the window cleaning robot to achieve an effective balance between coverage integrity and cleaning efficiency for different types of surfaces to be cleaned.

[0011] Optionally, a corresponding path planning strategy can be selected based on surface features, including:

[0012] If the surface width information is determined to be greater than the preset width threshold, the path planning strategy is determined to be the first strategy, and the main travel direction of the cleaning path corresponding to the first strategy is horizontal.

[0013] If the surface width information is determined to be less than or equal to the preset width threshold, and the boundary combination state is that both the left and right boundaries have borders, the path planning strategy is determined to be the second strategy, and the main travel direction of the cleaning path corresponding to the second strategy is longitudinal.

[0014] Thus, this application introduces a preset width threshold as a decision boundary point and performs secondary judgment based on the boundary combination state, enabling adaptive path direction guidance for different types of surfaces to be cleaned. For example, depending on whether the window is a "wide window" or a "narrow window," the main travel direction of the cleaning path (horizontal or vertical) is automatically switched to match the window geometry. In scenarios with narrow windows and frames on both sides, switching the main travel direction to vertical avoids frequent ineffective turns in the narrow horizontal space, thereby significantly improving cleaning efficiency. In wide window scenarios, using a horizontal main travel direction leverages the window's width to cover a larger area in a single pass, accelerating the overall cleaning speed and improving coverage speed in wide window scenarios.

[0015] Furthermore, this application enhances the targeting of path planning by using a dual judgment of preset width threshold and boundary combination state. For example, it can avoid misusing horizontal paths in narrow window scenarios or misusing vertical paths in wide window scenarios, ensuring that the path strategy is accurately matched with the window features.

[0016] Optionally, based on the selected path planning strategy, the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned, including:

[0017] When the first strategy is adopted, the window cleaning robot is controlled to move back and forth laterally from the starting position, and each time it reaches the first target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a lateral strip path covering the surface to be cleaned.

[0018] The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on a horizontal strip path.

[0019] Therefore, this application concretizes the first strategy into an execution process of lateral back-and-forth movement and boundary turning, achieving efficient coverage of the surface to be cleaned and ensuring the integrity of the coverage. For example, in wide-window scenarios, using lateral back-and-forth movement can maximize the coverage area of ​​a single trip and reduce the number of longitudinal movements, thereby significantly improving cleaning speed. Moreover, by performing a turning movement operation after each boundary is reached, it is ensured that there are no omissions between adjacent lateral strips, forming a continuous and complete coverage path.

[0020] Furthermore, by simplifying complex path planning into a loop of lateral movement and boundary turning, the implementation difficulty of the control logic is reduced, and the execution reliability is improved. This makes the first strategy particularly suitable for windows with large widths and regular boundaries, thus fully leveraging the efficiency advantages of lateral movement.

[0021] Optionally, the method also includes:

[0022] After the window cleaning robot finishes cleaning the surface to be cleaned based on the horizontal strip path, the robot is controlled to perform an edge cleaning strategy and return to the starting position.

[0023] Thus, this application incorporates an edge cleaning strategy after the main body cleaning and returns to the starting position. The main body cleaning (e.g., a horizontal strip path) primarily covers the central area of ​​the window, while the edges of the surface to be cleaned (especially near the four edges) are prone to cleaning blind spots due to path turning or boundary detection errors. The edge cleaning strategy can specifically target these edge areas for secondary cleaning, effectively eliminating residual dirt. Therefore, by designing a two-stage cleaning mode of main body cleaning and edge cleaning, comprehensive coverage from the center to the edges is achieved, significantly improving the overall cleaning effect. Furthermore, by controlling the window cleaning robot to return to the starting position from a known boundary location after edge cleaning, the risk of falling or efficiency reduction due to path uncertainty during the return trip is reduced.

[0024] In addition, for framed windows, dust and stains tend to accumulate near the edges. Edge cleaning strategies can target the edges of the windows, meeting users' higher requirements for window edge cleanliness.

[0025] Optionally, based on the selected path planning strategy, the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned, including:

[0026] When the second strategy is adopted, the window cleaning robot is controlled to move back and forth in the longitudinal direction, and each time it reaches the second target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a longitudinal strip path covering the surface to be cleaned;

[0027] The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on a longitudinal strip path.

[0028] Thus, this application improves the cleaning efficiency and coverage of the surface to be cleaned by specifying the second strategy as a longitudinal back-and-forth and boundary-turning execution process. It is particularly suitable for scenarios with narrow windows and frames on both sides. In this scenario, using longitudinal back-and-forth movement can avoid frequent and ineffective turns in the narrow lateral space, thereby significantly improving cleaning efficiency. Moreover, the longitudinal strip path can better conform to the geometry of the narrow window, ensuring no omissions in the coverage of the window surface with limited width. Compared to lateral back-and-forth, longitudinal back-and-forth can take advantage of the height of the window surface, allowing a longer distance to be covered in a single trip, reducing the number of turns, and reducing the time loss and path overlap caused by turns.

[0029] Furthermore, in the scenario described above where the window is narrow and has frames on both sides, the second strategy, by using longitudinal movement, allows the window cleaning robot to be closer to the frame on both sides, and the turning action is performed laterally, which helps to reduce the risk of falling due to excessive turning.

[0030] Optionally, before controlling the window cleaning robot to move back and forth longitudinally, the method further includes:

[0031] Control the window cleaning robot to perform a lateral edge probing and width measurement operation to obtain the available lateral width of the surface to be cleaned;

[0032] After obtaining the available horizontal width, the body posture of the window cleaning robot is adjusted;

[0033] If the robot body is found to be obstructed during the posture adjustment process, the robot body is controlled to retreat longitudinally by a preset safe distance, and the robot body posture is readjusted again so that the front side of the robot body contacts the boundary of the surface to be cleaned.

[0034] Based on the type of signal generated when the front side of the fuselage contacts the boundary, determine whether there is a restricted area on the surface to be cleaned;

[0035] After determining that there are restricted areas on the surface to be cleaned, the number of strips for the longitudinal strip path is determined based on the available lateral width.

[0036] Thus, this application improves path planning accuracy and enhances the robustness of robot posture adjustment by adding lateral edge probing, attitude adjustment, and restricted area judgment before longitudinal cleaning to determine the number of strips in the longitudinal strip path. For example, by performing lateral edge probing and width measurement, the actual usable width of the window surface can be accurately obtained, providing accurate data for subsequent calculation of the number of longitudinal strips and avoiding coverage omissions or path overlaps caused by width estimation errors. By detecting steering obstruction and performing a reversal operation, the problem of the robot's inability to complete attitude adjustment in narrow spaces is solved, ensuring that the window cleaning robot can smoothly enter the longitudinal cleaning mode.

[0037] Furthermore, by analyzing the signal type when the fuselage's front side contacts the boundary, this application can distinguish between two different scenarios: narrow-area confinement and corner confinement, providing more refined environmental information for subsequent path planning. Based on the available lateral width and confined area information, the number of longitudinal stripes is dynamically determined, ensuring reasonable coverage density within a limited width. This avoids efficiency degradation due to too many stripes or omissions due to too few strips, thus optimizing the coverage strategy for longitudinal strip paths.

[0038] Optionally, a corresponding path planning strategy can be selected based on surface features, including:

[0039] Obtain the current task type of the window cleaning robot, and select the corresponding path planning strategy based on the current task type and surface features.

[0040] In this way, by introducing the current task type as an additional decision-making dimension for path planning strategy selection, this application enables the window cleaning robot to adjust its path strategy according to the specific task to be completed, rather than using the same path pattern for all tasks, thus achieving task-oriented differentiated paths. For example, in the edge cleaning task, the window cleaning robot can directly enter the edge cleaning path without first executing the global lateral strip path, thereby avoiding unnecessary path waste and improving task execution efficiency.

[0041] Furthermore, when users select different task types based on their actual needs, the window cleaning robot can automatically adapt its path strategy according to the selected task type, thus improving the product's ease of use and intelligence.

[0042] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0043] If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that at least one of the left and right boundaries has a border, control the window cleaning robot to execute the edge cleaning strategy so that the window cleaning robot moves and cleans along the boundary of the surface to be cleaned.

[0044] If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that neither the left nor the right boundary has a border, or the surface width information is less than or equal to a preset threshold, the window cleaning robot is controlled to execute a frameless return-to-start strategy so that the window cleaning robot returns to the starting position.

[0045] Thus, this application differentiates the edge cleaning task in framed and frameless scenarios by selecting different path planning strategies. This not only improves the targeting and safety of edge cleaning but also avoids ineffective edge following in frameless scenarios, optimizing task execution efficiency. Specifically, in scenarios with double-sided frameless or extremely narrow windows, executing an edge cleaning strategy carries the risk of falling or failing to form a closed-loop path. By switching to a frameless return-to-start strategy, the window cleaning robot avoids ineffective or dangerous boundary following at frameless edges. Furthermore, performing edge cleaning when there is at least one framed boundary ensures that the window cleaning robot has a reliable physical boundary to follow, thereby improving the safety and effectiveness of edge cleaning.

[0046] In addition, this application identifies extremely narrow window surfaces by setting a preset threshold, avoiding edge-to-edge operations on window surfaces that are not wide enough to support safe edge-to-edge movement, reducing the risk of falls or collisions, and improving the robustness of the window cleaning robot in performing tasks.

[0047] It should also be noted that in frameless or extremely narrow scenarios, controlling the window cleaning robot to return directly to the starting point instead of attempting to perform edge cleaning can also avoid long periods of ineffective movement or getting stuck due to the inability to complete the edge path, thus improving the overall task execution efficiency.

[0048] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0049] If the current task type is determined to be a global cleaning task, and the boundary combination state is determined to be that both the left and right boundaries have borders, and the surface width information is greater than a preset threshold, the window cleaning robot is controlled to execute a full-coverage path strategy.

[0050] If the current task type is determined to be a global cleaning task, and the surface features are determined to meet at least one of the following conditions: the boundary combination state is that at least one of the left and right boundaries has no border, the surface width information is less than or equal to a preset threshold, or no surface width information is detected, then the window cleaning robot is controlled to execute a protection strategy. The protection strategy includes at least one of the following: a backtracking strategy, a return to the starting point strategy, or termination of the cleaning operation.

[0051] Therefore, if a window cleaning robot performs a global cleaning task on a frameless, extremely narrow, or surface with incomplete width information, there is a very high risk of it falling. Thus, by assessing the safety of the surface to be cleaned during a global cleaning task and implementing protective strategies for unsafe scenarios, the window cleaning robot effectively avoids accidents and prevents global cleaning from being performed on dangerous surfaces. Furthermore, by proactively identifying unsafe surfaces and taking protective measures, the risk of the window cleaning robot falling and being damaged is reduced.

[0052] Moreover, the protection strategy provides multiple options (reverse, return to starting point, end operation), allowing the window cleaning robot to choose a safe response method according to the specific scenario, thus enhancing the safety redundancy and robustness of the window cleaning robot.

[0053] It should also be noted that when surface features do not meet global cleaning requirements (such as extremely narrow windows), forcibly implementing a full-coverage path strategy may lead to path overlap, inefficiency, or omissions in coverage. This application addresses the situation where surface features do not meet global cleaning requirements by implementing a protection strategy to avoid invalid cleaning operations.

[0054] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0055] If the current task type is determined to be a local cleaning task, and the surface width information of the local area is determined to meet the preset conditions, the window cleaning robot is controlled to execute the first strategy.

[0056] If the current task type is determined to be a local cleaning task, and the surface width information of the local area does not meet the preset conditions, or the boundary combination state of the local area is that at least one of the left and right boundaries does not have a border, the window cleaning robot is controlled to execute a local strip coverage strategy so that the window cleaning robot can move back and forth to clean the local area based on the strip coverage path.

[0057] In this way, by selecting different path strategies based on the width and boundary conditions of the local area during local cleaning tasks, fine-grained path adaptation for local areas is achieved, avoiding the efficiency reduction or safety risks caused by misusing lateral paths in narrow windows or frameless local areas. Furthermore, for narrow windows or frameless local areas, the local strip coverage strategy ensures complete coverage without omissions, improving the overall coverage integrity of local cleaning. For wide windows with frames on both sides, the first strategy is used to fully utilize the width advantage, quickly completing the cleaning of the local area and optimizing the efficiency of local cleaning.

[0058] In addition, for frameless or incomplete boundary areas, a local strip coverage strategy can be adopted to prevent the magnetic storage robot from approaching the frameless edge due to lateral movement, reduce the risk of falling, and enhance the safety of local cleaning.

[0059] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0060] If the current task type is determined to be a zone cleaning task, the surface to be cleaned is divided into multiple target areas;

[0061] Select the corresponding path planning strategy based on the surface width and boundary type information of each target area.

[0062] Thus, this application achieves refined path planning for complex surfaces by dividing the surface to be cleaned into multiple target areas and adapting each area to a specific path planning strategy. This is particularly beneficial for complex window surfaces with irregular shapes, varying widths, or inconsistent boundary types, where a single path strategy cannot cover all areas. The zoned cleaning task allows for the selection of an appropriate path planning strategy for each area, enabling refined path planning.

[0063] Furthermore, by dividing the surface to be cleaned into multiple target areas and implementing corresponding path planning strategies for each, overall cleaning efficiency can be improved. For example, an efficient lateral back-and-forth strategy can be used in wide window areas, while a vertical strip strategy can be used in narrow window areas, avoiding efficiency losses caused by using a single strategy globally. For irregularly shaped windows such as L-shaped, T-shaped, and curved surfaces, the zoned cleaning task can divide the surface into multiple regular sub-areas, with each sub-area employing a corresponding path planning strategy, thereby achieving effective coverage of irregularly shaped windows.

[0064] In addition, by rationally dividing areas and connecting paths between areas, a smooth transition of cleaning paths between different areas can be ensured, reducing duplicate coverage or omissions, thereby optimizing the continuity of cleaning paths.

[0065] Optionally, the method also includes:

[0066] After the window cleaning robot has completed any of the tasks of zone cleaning, global cleaning, or edge cleaning, obtain the cleaning mode and boundary type information of the surface to be cleaned.

[0067] Based on cleaning mode and boundary type information, determine whether the window cleaning robot should perform an edge supplementary cleaning strategy.

[0068] Thus, this application determines whether the window cleaning robot should execute the edge-supplementary cleaning strategy by introducing dual judgment of cleaning mode and boundary type information after the completion of any of the partition cleaning task, global cleaning task, or edge-supplementary cleaning task. This avoids ineffective edge-supplementary cleaning in frameless scenarios and optimizes the total task time. For example, in frameless boundary scenarios, even in deep cleaning mode, the edge-supplementary cleaning strategy is skipped, preventing the window cleaning robot from performing dangerous or ineffective boundary following at frameless edges. In fast cleaning mode, skipping the edge-supplementary cleaning strategy can significantly shorten the total task time, meeting users' needs for rapid cleaning.

[0069] Moreover, based on the above control logic, users do not need to manually select whether to execute the edge supplement cleaning strategy. Instead, the window cleaning robot automatically makes decisions based on the cleaning mode and boundary type, which simplifies user operation and improves the flexibility of user experience.

[0070] Furthermore, after users select a cleaning mode according to their actual needs, the window cleaning robot can adjust the cleaning quality as required. Specifically, in deep cleaning mode, it automatically performs a supplementary edge cleaning strategy to ensure no blind spots at the edges; in quick cleaning mode, it skips the edge cleaning strategy to improve cleaning efficiency.

[0071] Optionally, the method also includes:

[0072] If it is determined that the window cleaning robot will perform a supplementary cleaning strategy along the edges, control the window cleaning robot to perform the following operations:

[0073] Determine the starting boundary for edge-to-edge cleaning based on boundary type information;

[0074] The window cleaning robot is controlled to move sequentially along the starting boundary in a first preset direction to perform supplementary cleaning on the edge areas of the surface to be cleaned;

[0075] Record the number of times the window cleaning robot cleans along the edges, and when the number of times it cleans along the edges reaches the preset number, control the window cleaning robot to execute the return-to-start strategy.

[0076] In this way, by controlling the window cleaning robot to move along the edges and recording the number of times it moves, the system ensures that the edge areas are cleaned at least a preset number of times, avoiding residual dirt due to insufficient cleaning and guaranteeing thorough cleaning of the edge areas. Furthermore, by clearly defining the starting boundary and the first preset direction, the edge cleaning path becomes more standardized and predictable, reducing path confusion or repeated coverage and improving the standardization of edge cleaning paths. Moreover, by setting an automatic return-to-starting-point strategy after completing supplementary edge cleaning, the system ensures the window cleaning robot safely returns to its starting position, avoiding inconvenience for users retrieving the robot due to uncertain task completion locations.

[0077] Furthermore, this application achieves automated control of edge cleaning by recording the number of times the window cleaning robot performs supplementary edge cleaning and judging the preset number of edge cleaning cycles, without requiring user intervention and improving the user experience.

[0078] Optionally, the method also includes:

[0079] Obtain the number of edge loops and / or edge entry conditions corresponding to the execution of the edge supplementary cleaning strategy;

[0080] The window cleaning robot is controlled to perform a supplementary cleaning strategy along the edges based on the number of edge loops and / or edge entry conditions.

[0081] In this way, by setting the number of cleaning cycles along the edge, users or window cleaning robots can flexibly adjust the intensity of edge cleaning according to their cleaning needs, achieving precise control over edge cleaning and thus balancing cleaning effectiveness and time consumption. Furthermore, by setting the edge entry conditions, different triggering conditions can be met, making the triggering of supplementary edge cleaning more intelligent and flexible. Moreover, by judging the edge entry conditions, this step can be skipped in scenarios where edge cleaning is not needed, avoiding unnecessary edge cleaning.

[0082] In addition, users can set the number of rings along the edge and entry conditions to achieve personalized cleaning configurations and improve the user experience.

[0083] Optionally, obtain the surface features of the surface to be cleaned by the window cleaning robot, including:

[0084] The window cleaning robot is controlled to move along a second preset direction to the third target boundary of the surface to be cleaned, and performs edge probing and edge-adhering actions to obtain the surface features of the surface to be cleaned by the window cleaning robot.

[0085] In this way, by controlling the window cleaning robot to perform edge probing actions, the boundary position and type can be accurately detected, avoiding feature information deviations caused by sensor errors or irregularities in the surface to be cleaned, thus improving the accuracy of surface feature acquisition. By controlling the window cleaning robot to perform edge-fitting actions, the robot body is made to fit tightly against the boundary, ensuring the accuracy of the starting position of the subsequent cleaning path, avoiding path deviations or omissions caused by starting position deviations, and ensuring the accuracy of the starting position of path planning. Therefore, using edge probing and edge-fitting actions as important preliminary steps in path planning, by obtaining accurate boundary information, provides accurate input parameters for subsequent path strategy selection.

[0086] Furthermore, for windows with complex boundary types (such as partially framed and partially frameless) or irregular shapes, more accurate local feature information can be obtained by actively performing edge probing and edge-fitting actions, thereby enhancing the adaptability to complex surfaces to be cleaned.

[0087] Optionally, the method also includes:

[0088] During the cleaning operation of the window cleaning robot on the surface to be cleaned, if an interruption is detected, the breakpoint information is recorded. The breakpoint information includes at least the surface width information corresponding to the surface to be cleaned that the window cleaning robot was adsorbed at the time of the interruption.

[0089] When it is determined that the window cleaning robot has resumed cleaning operations, the current surface characteristics of the surface to be cleaned by the window cleaning robot are reacquired.

[0090] Based on the current surface features and breakpoint information, determine whether to execute the breakpoint continuation cleaning strategy.

[0091] In this way, by recording the breakpoint information when the window cleaning robot is interrupted and intelligently matching it upon resumption, the robot can continue cleaning from the breakpoint position after the interruption is resolved, instead of starting from the beginning. This avoids repeated coverage of already cleaned areas and significantly improves cleaning efficiency. Furthermore, users do not need to manually replan or wait for the robot to re-clean the entire surface after an interruption, reducing user waiting time and operational steps, and enhancing the product's usability and intelligence.

[0092] Therefore, when a user removes the window cleaning robot from one surface to be cleaned and places it on another, the robot can automatically identify whether the surface has changed by using width matching. This allows it to decide whether to execute the breakpoint resume cleaning strategy, avoiding erroneous resume cleaning on different surfaces. Furthermore, by recording breakpoint information, even in the event of an unexpected interruption (such as a power outage or a fall), the robot can quickly locate and continue its task after recovery, enhancing its ability to handle abnormal situations.

[0093] Optionally, the breakpoint information also includes the boundary type information and boundary combination state of the surface to be cleaned by the window cleaning robot when the interruption occurs; based on the current surface features and breakpoint information, it is determined whether to execute the breakpoint continuation cleaning strategy, including:

[0094] If the current surface features and breakpoint information meet preset judgment conditions, the window cleaning robot is controlled to execute a breakpoint continuation cleaning strategy; the preset judgment conditions include at least one of the following:

[0095] The difference between the current surface width information and the surface width information in the breakpoint information is less than the preset difference threshold;

[0096] The current surface width is greater than or equal to the preset minimum width threshold;

[0097] The boundary type information and boundary combination status are consistent with the current boundary type information and current boundary combination status in the current surface features.

[0098] Thus, by introducing breakpoint information and more refined judgment conditions—namely, by using current surface width information, boundary type information, and boundary combination state—this application can more accurately determine whether the window cleaning robot is still on the same surface or in the same area to be cleaned, avoiding misjudgments caused by coincidental width matching and improving the accuracy of resuming cleaning after a breakpoint. Furthermore, the preset judgment conditions support the execution of the resuming cleaning strategy as long as at least one condition is met, allowing the window cleaning robot to select the appropriate judgment logic according to the actual scenario, thereby providing flexible judgment logic.

[0099] It should also be noted that when a user removes a window cleaning robot from one surface to be cleaned and places it on another, even if the widths happen to be the same, the boundary type information and boundary combination state may differ. By matching the aforementioned preset judgment conditions, it is possible to accurately identify whether the surface to be cleaned has been changed, avoiding the incorrect execution of continued cleaning on different surfaces.

[0100] Furthermore, by setting the current surface width information to be greater than or equal to the preset minimum width threshold, the risk of performing follow-up cleaning on extremely narrow surfaces to be cleaned can be eliminated, because extremely narrow surfaces to be cleaned may not be able to safely perform follow-up cleaning paths.

[0101] Optionally, the method also includes:

[0102] If the current surface features and breakpoint information do not meet the preset judgment conditions, the corresponding path planning strategy is selected based on the current surface features, and the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

[0103] Thus, if a pre-defined judgment condition determines that the surface to be cleaned has been replaced or its characteristics have changed, forcibly executing a breakpoint-based cleaning strategy may lead to path errors, duplicate coverage, or omissions. By abandoning the breakpoint-based cleaning strategy and reselecting a path planning strategy, efficiency losses caused by incorrect cleaning continuation are avoided. Moreover, in cases where cleaning cannot be continued, reselecting a path planning strategy and executing the cleaning operation ensures that the current surface to be cleaned is still completely cleaned, preventing uncleaned areas from being left due to interruption and ensuring the integrity of the cleaning task.

[0104] Furthermore, because the window cleaning robot can automatically adjust its path planning strategy based on the current surface characteristics, no user intervention is required. Even if the surface conditions change after an interruption, the window cleaning robot can quickly adapt and start a new cleaning task, enhancing its adaptability. Moreover, the above process eliminates the need for users to manually determine whether to execute a resume cleaning strategy or restart the path planning strategy; the window cleaning robot automatically completes the decision-making and switching, simplifying user operation and improving the product's intelligence level.

[0105] Optionally, the breakpoint information also includes the window cleaning robot's position information at the time of the interruption, controlling the window cleaning robot to execute the breakpoint continuation cleaning strategy, including:

[0106] Starting with the location information, the window cleaning robot continues to execute the corresponding path planning strategy to perform cleaning operations on the surface to be cleaned.

[0107] In this way, by utilizing location information to precisely resume cleaning from the breakpoint, the window cleaning robot can accurately return to the position where the interruption occurred, rather than restarting from the beginning of the entire surface to be cleaned. This minimizes the area of ​​repeated cleaning and achieves precise resume cleaning from the breakpoint. Since cleaning only needs to continue from the breakpoint, instead of re-cleaning the entire surface, cleaning time can be significantly saved, especially when the surface to be cleaned is large or most of the area has already been cleaned, thus improving cleaning efficiency.

[0108] Furthermore, by using location information for path connection processing, the continuity and integrity of the restored path are ensured to be consistent with the path before the interruption, avoiding coverage omissions or duplications caused by path misalignment.

[0109] Furthermore, the above process does not require users to manually record the interruption location or replan the path. The window cleaning robot can automatically complete the location positioning and path restoration, which greatly improves the ease of use and intelligent experience of the product.

[0110] Optionally, the method also includes:

[0111] During the cleaning operation of the window cleaning robot on the surface to be cleaned, the feedback information of various sensors in the window cleaning robot is detected;

[0112] When feedback indicates that the window cleaning robot has encountered an anomaly or boundary anomaly, the robot is controlled to correct its path planning strategy so that it can perform cleaning operations based on the corrected path planning strategy.

[0113] In this way, this application improves the safety of the cleaning process by monitoring the window cleaning robot in real time for any abnormalities, promptly correcting its path, or executing safety actions to prevent damage or accidents. Furthermore, by adjusting the path planning strategy in real time, the window cleaning robot can cope with various emergencies, reducing cleaning interruptions or task failures caused by abnormalities and improving task completion rates. This dynamic correction mechanism makes the path planning strategy more robust, adapting to changes in the conditions of different types of surfaces to be cleaned, ensuring complete cleaning coverage.

[0114] Furthermore, when encountering boundary anomalies (such as irregular window shapes or abrupt changes in boundary types), the window cleaning robot can continue to complete the cleaning task by dynamically correcting the path planning strategy, instead of stopping or failing directly, thus enhancing the adaptability of the window cleaning robot to complex surfaces to be cleaned.

[0115] Optionally, the window cleaning robot performs cleaning operations based on a modified path planning strategy, including any of the following:

[0116] Control the window cleaning robot to pause its cleaning operation;

[0117] After controlling the window cleaning robot to retreat a preset distance, it performs the cleaning operation again based on the path planning strategy;

[0118] After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy.

[0119] Control the window cleaning robot to execute the edge cleaning strategy.

[0120] Different abnormal situations require different responses. Therefore, this application provides four correction methods: pause, reverse, path switching, and edge cleaning. This allows the window cleaning robot to select the appropriate correction action based on the type of abnormality, thus improving its flexibility in responding to situations.

[0121] Among these improvements, the corrective methods of retrying or switching paths allow the window cleaning robot to bypass minor obstacles or abnormal areas and continue cleaning, rather than stopping or failing directly, thus improving the robustness of the cleaning task. The corrective method of switching to an edge cleaning strategy ensures that edge areas are still cleaned even in cases of boundary abnormalities, preventing edge omissions and ensuring thorough edge coverage. The corrective method of pausing execution allows the window cleaning robot to stop promptly when encountering serious abnormalities (such as insufficient suction or sensor failure), preventing damage to the robot or safety accidents and improving operational safety.

[0122] Optionally, the method also includes:

[0123] During the cleaning operation of the window cleaning robot on the surface to be cleaned, if the window cleaning robot encounters abnormal working conditions, the type of abnormality is determined.

[0124] Based on the type of exception, the window cleaning robot selects the corresponding processing strategy so that the window cleaning robot can perform the corresponding operation based on the processing strategy.

[0125] Thus, by introducing anomaly type classification and differentiated handling strategies, this application not only achieves precise anomaly response but also improves the efficiency of anomaly handling. Different anomaly types require different handling methods. Therefore, by identifying the anomaly type, appropriate handling strategies can be selected specifically, improving the accuracy of the response. Furthermore, for different types of anomalies, the window cleaning robot can quickly match preset handling strategies without complex comprehensive judgments, improving the response speed of anomaly handling. Moreover, through differentiated handling strategies, most anomalies can be automatically handled without disturbing the user, improving the user experience.

[0126] Furthermore, the design of the aforementioned differentiated handling strategies enhances the window cleaning robot's adaptability to various abnormal situations, thereby improving its robustness and safety. For example, in the case of safety-related anomalies, the window cleaning robot can prioritize the execution of safety handling strategies to ensure robot safety; in the case of mechanical anomalies, the window cleaning robot can attempt to resume operation to avoid task interruption.

[0127] Optionally, the window cleaning robot performs corresponding operations based on the processing strategy, including any of the following:

[0128] Control the window cleaning robot to perform local obstacle avoidance actions, and after obstacle avoidance, continue to perform cleaning operations based on the selected path planning strategy;

[0129] After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy.

[0130] Control the window cleaning robot to prevent it from continuing the breakpoint cleaning strategy;

[0131] Control the window cleaning robot to pause its cleaning operation;

[0132] Control the window cleaning robot to execute the return-to-start strategy.

[0133] Different types and severity of anomalies require different actions. Therefore, this application provides five methods—local obstacle avoidance, path switching, disabling interrupted cleaning, pausing execution, and returning to the starting point—to enable the window cleaning robot to select the appropriate operation based on the anomaly situation, thereby improving its flexibility in handling anomalies. Specifically, by using local obstacle avoidance or path switching, the window cleaning robot can bypass minor anomalies and continue cleaning, avoiding task interruption and improving the continuity of the cleaning task.

[0134] By prohibiting breakpoint-based resume cleaning, the window cleaning robot can avoid erroneous resume cleaning after the surface to be cleaned is changed or the sensor fails, which could lead to path confusion or missed coverage, effectively preventing incorrect resume cleaning. The pause or return-to-start strategy allows the window cleaning robot to stop or return promptly when encountering serious abnormalities (such as insufficient suction or risk of falling), ensuring the robot's safety.

[0135] In this way, the window cleaning robot can handle minor anomalies automatically without the user noticing; for serious anomalies, the window cleaning robot can notify the user by pausing execution or returning to the starting point, thereby optimizing the user experience.

[0136] Optionally, the method also includes:

[0137] After completing the cleaning operation, the corresponding return-to-start strategy is selected based on the path planning strategy and / or surface characteristics, and the window cleaning robot is controlled to return to the starting position based on the return-to-start strategy.

[0138] Thus, this application, by intelligently selecting a return-to-start strategy based on path planning and / or surface characteristics, not only improves the efficiency of the return-to-start process but also ensures its safety. Specifically, by selecting an appropriate return-to-start strategy, the return path can be shortened, return time reduced, and overall cleaning efficiency improved. Selecting a return-to-start strategy based on surface characteristics avoids the risk of collisions or falls due to improper path selection on narrow or frameless surfaces, ensuring a safe return process.

[0139] Furthermore, by associating the return-to-start strategy with the path planning strategy, this application can ensure that the return path is consistent with the cleaning path, avoiding path conflicts or confusion, optimizing the integrity of path planning, and through intelligent selection of the return-to-start strategy, the window cleaning robot can automatically, efficiently and safely return to the starting point without manual intervention from the user, thus improving the product's intelligence level and user experience.

[0140] Optionally, a corresponding return-to-start strategy can be selected based on the path planning strategy and / or surface features, including:

[0141] When the path planning strategy is a strategy with the longitudinal direction as the main direction of travel, the window cleaning robot is controlled to return to the starting point along the longitudinal direction.

[0142] When the path planning strategy is a strategy with the lateral direction as the main travel direction, the general strategy for controlling the window cleaning robot is to return to the starting point.

[0143] When the surface features indicate that the boundary of the surface to be cleaned is incomplete, the window cleaning robot adopts a frameless return-to-start strategy.

[0144] When the path planning strategy is an edge cleaning strategy, the current boundary and starting edge information are determined based on the surface features. After adjusting the body posture of the window cleaning robot based on the current boundary and starting edge information, the window cleaning robot is controlled to adopt a general return-to-starting-point strategy.

[0145] Different cleaning paths and surface conditions require different return-to-start paths. Therefore, this application achieves differentiated return-to-start path planning by intelligently selecting different return-to-start strategies based on path planning strategies and surface characteristics. This allows the window cleaning robot to choose the appropriate return method according to the actual scenario. For example, by controlling the window cleaning robot to return to the starting point in the longitudinal direction, unnecessary lateral switching actions can be avoided in narrow window scenarios, reducing the return path length and time, and improving return efficiency.

[0146] By controlling the window cleaning robot to adopt a universal return-to-start strategy, the window cleaning robot can ensure the integrity and accuracy of the return path in wide window scenarios through a three-stage return process of "vertical retreat, lateral correction, and compensation when necessary".

[0147] By controlling the window cleaning robot to adopt a frameless return-to-start strategy, the window cleaning robot can use a more conservative "edge probe-reverse-turn-re-edge probe" method, avoiding the risk of falling due to reliance on physical frames and ensuring the safety of the frameless window surface during the return trip.

[0148] By controlling the window cleaning robot to adopt a dedicated return-to-origin strategy after cleaning along the edge, and by adjusting the posture based on the starting and ending postures of the edge cleaning, a smooth transition from edge cleaning to returning to the origin is ensured, avoiding path confusion caused by posture conflicts of the window cleaning robot.

[0149] Therefore, this application can select differentiated return-to-origin strategies based on different surface characteristics of the surface to be cleaned and / or different path planning strategies. For example, different return methods can be used for narrow windows, frameless windows, and different main path patterns. Compared with the unified return-to-origin scheme in the prior art, this application can reduce redundant paths and boundary risks in the return process, and improve the success rate, stability, and execution efficiency of the return-to-origin action in complex scenarios.

[0150] Secondly, this application provides a control device for a window cleaning robot, the device comprising:

[0151] The acquisition module is used to acquire the surface features of the surface to be cleaned by the window cleaning robot. The surface features include surface width information, boundary type information, and boundary combination state.

[0152] The control module is used to select the corresponding path planning strategy according to the surface characteristics, and control the window cleaning robot to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

[0153] Thirdly, this application provides a window cleaning robot, including: a memory and a processor;

[0154] The memory stores instructions that the computer executes;

[0155] The processor executes computer execution instructions stored in memory, causing the processor to perform the method as described in any of the first aspects.

[0156] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any of the first aspects.

[0157] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in any of the first aspects.

[0158] It should be noted that the second to fifth aspects of this application have similar beneficial effects to the corresponding technical solutions in the first aspect of this application, and the corresponding feasible implementation methods will not be repeated here.

[0159] The control method, device, robot, medium, and product for a window cleaning robot provided in this application acquire surface features of the surface to be cleaned by the robot. These surface features include surface width information, boundary type information, and boundary combination state. The surface width information characterizes the lateral or longitudinal dimensions of the surface to be cleaned; the boundary type information distinguishes between framed boundaries, frameless boundaries, virtual boundaries, or mixed boundaries; and the boundary combination state describes the relationship between the left and right boundaries, such as framed on both sides, framed on one side, or frameless on both sides. Compared to existing technologies that only distinguish between framed and frameless windows with coarse-grained modeling, this application achieves refined identification and quantitative characterization of window features. Furthermore, based on the acquired surface features, a matching path planning strategy is selected. By recognizing multi-dimensional features such as surface width, boundary type, and boundary combination states, this technology can automatically match appropriate path planning strategies for different types of surfaces to be cleaned (e.g., narrow windows, wide windows, framed windows, frameless windows, mixed boundary windows, etc.), significantly improving the adaptability of the window cleaning robot under complex surface conditions and overcoming the path pattern mismatch problem caused by existing technologies that only distinguish between framed and frameless windows. Furthermore, based on the selected path planning strategy, the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned. Since the selected path planning strategy is closely related to the characteristics of the surface to be cleaned, it can effectively avoid problems such as repeated wiping and partial omissions caused by path pattern mismatch with window type, ensuring full coverage cleaning of the surface and improving cleaning quality. Moreover, this application can adapt to scenarios with inconsistent boundary shapes, such as one side being framed and the other side being frameless, or situations with mixed boundaries in local areas. By dynamically adjusting the path planning strategy, it ensures stable execution of cleaning tasks under different boundary combination states, enhancing the robustness and reliability of the window cleaning robot in practical use. Therefore, this application reduces ineffective turning, redundant paths, and repetitive cleaning behaviors of window cleaning robots by refining surface feature recognition and dynamic path selection. This is particularly effective in scenarios with narrow windows or asymmetrical boundary conditions, significantly shortening cleaning time and improving overall operational efficiency. Furthermore, the process does not rely on preset templates or manual intervention, enhancing the window cleaning robot's adaptability to different types of surfaces. Attached Figure Description

[0160] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0161] Figure 1 This is a schematic diagram of an application scenario provided by an embodiment of this application;

[0162] Figure 2A flowchart illustrating a control method for a window cleaning robot provided in an embodiment of this application;

[0163] Figure 3 A flowchart illustrating the differentiated return-to-start mechanism of a window cleaning robot provided in this application embodiment;

[0164] Figure 4 A schematic diagram of the overall process for full-coverage path planning of a window cleaning robot based on window type differences, provided in an embodiment of this application;

[0165] Figure 5 A flowchart illustrating the selection of a window type recognition and path planning strategy is provided in an embodiment of this application.

[0166] Figure 6 A flowchart for breakpoint continuation cleaning and window matching determination is provided in the embodiments of this application;

[0167] Figure 7 This is a schematic diagram of the control device for a window cleaning robot provided in an embodiment of this application;

[0168] Figure 8 This is a schematic diagram of the structure of a window cleaning robot provided in an embodiment of this application.

[0169] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0170] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0171] To facilitate a clear description of the technical solutions in the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish identical or similar items with essentially the same function and effect. For example, the first strategy and the second strategy are only used to distinguish different strategies and do not limit their order. 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.

[0172] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0173] In this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0174] Existing window cleaning robots, especially window wiping robots, typically employ a template-based, full-coverage cleaning solution based on edge detection for path planning. This type of solution generally uses collision sensors, pressure sensors, optical sensors, laser sensors, or attitude sensors to detect window edges, framed or frameless edges, and then combines this with a pre-set motion template to achieve a reciprocating cleaning motion on the target window surface. Common path types include zigzag paths, N-shaped paths, horizontal or vertical zigzag paths, and paths that perform edge-following cleaning and return to the starting point after completing the main cleaning.

[0175] However, in existing technologies, most products generally use edge detection and templated coverage paths, combined with local correction or return to the starting point for window cleaning. Some products also provide different cleaning modes such as deep cleaning, fast cleaning, edge cleaning, and spot cleaning, and make local motion adjustments based on edge detection results during operation to reduce the risk of falls and improve adaptability to common window scenarios such as framed and frameless windows.

[0176] In addition, some products have introduced features such as resume cleaning after interruption, location memory, or path memory to improve the user experience. However, these solutions mainly rely on fixed path templates and achieve full window coverage cleaning through boundary detection and simple rule switching.

[0177] Although existing window cleaning robots can automatically clean most regular window surfaces, they still have the following obvious shortcomings when faced with different window types, different boundary combinations, and different operation interruption scenarios:

[0178] First, most solutions only distinguish between framed and frameless windows, or only perform avoidance, turning, or stopping actions when an edge is detected, lacking further identification and utilization of factors such as window width, left and right frame combination, local area range, narrow window, wide window difference, and mixed boundary conditions.

[0179] While existing technologies can perform edge detection, they typically cannot make fine-grained adaptive adjustments to the global path pattern based on window type differences. Therefore, in narrow windows, narrow windows with frames on both sides, locally partitioned windows, or windows with inconsistent boundary shapes, there is often a lack of a unified and effective differentiated planning mechanism. This can easily lead to a mismatch between the cleaning path and the actual window type, resulting in problems such as repeated wiping, edge omissions, invalid turns, low local cleaning efficiency, and insufficient overall coverage integrity.

[0180] Secondly, existing technologies generally use preset Z-shaped, N-shaped, or similar back-and-forth templates as the main cleaning path, lacking a systematic design for switching rules of path modes under different window conditions. For example, most existing solutions use fixed template paths such as "zigzag," "back-and-forth," and "Z / N cleaning," but rarely use methods that dynamically determine the path mode based on window characteristics such as measured window width, border presence, and boundary combination relationships. This leads to a situation where, when the window width is close to the body size, local areas are narrow, or boundary conditions are asymmetrical, fixed template paths struggle to balance full coverage and cleaning efficiency, and may even increase the probability of invalid turns, repeated paths, and edge omissions.

[0181] Furthermore, existing cleaning methods for resuming cleaning after a breakpoint typically focus on position or path memory, but lack a mechanism to identify whether the current window surface is the same as the one before the interruption. In other words, when resuming cleaning after a breakpoint, existing technologies often assume that the current resumed environment is consistent with the original interrupted environment, without fully considering that the user may reposition the robot to another window, another piece of glass, or a different area of ​​the same window surface.

[0182] Therefore, if there is a lack of joint judgment on window type information such as window width, border distribution, and boundary features, and the system only relies on historical position or historical mode to continue execution, there is a lack of a judgment mechanism to determine whether the current window surface is consistent with the window surface corresponding to the original breakpoint. This can easily cause misalignment of the cleaning continuation, incorrect recovery, abnormal local coverage, inapplicability of the original cleaning strategy, or direct skipping of the area to be cleaned after the user repositions the device, thus affecting the full coverage cleaning effect.

[0183] Furthermore, existing technologies often employ a return-to-starting-point strategy or a return-to-origin strategy when window cleaning robots finish cleaning or encounter abnormal situations, without tailoring the design to different window types or path patterns. For scenarios such as narrow windows, frameless windows, edge-following patterns, and local area patterns, using a uniform return-to-origin strategy can easily lead to problems such as lengthy return paths, insufficient return stability, increased risks at local edges, or return failures. Especially in frameless or narrow window scenarios, if the return action is not specifically planned based on the current window characteristics, its stability and execution efficiency are often difficult to guarantee, resulting in return-to-origin failures or low return efficiency.

[0184] In summary, the root cause of the aforementioned problems lies in the fact that existing path planning technologies primarily employ a paradigm that combines fixed template paths with edge-triggered path correction. Furthermore, the edge detection corresponding to edge-triggered path correction is mostly used for security protection, local correction, or simple mode switching, rather than being further transformed into window-type feature variables that can participate in global decision-making, nor has a path planning decision-making chain based on window-type differences been formed.

[0185] In other words, existing technologies lack a comprehensive path planning mechanism that addresses window type differences. They cannot integrate window width, boundary existence, boundary combination relationships, regional attributes, interruption information, and return strategies into a unified planning framework for collaborative decision-making. Therefore, under complex window types or multi-scenario switching conditions, it is difficult to simultaneously ensure coverage integrity, path rationality, recovery accuracy, and execution efficiency.

[0186] To address the aforementioned issues, this application provides a control method for a window cleaning robot. This method acquires the surface features of the surface to be cleaned by the robot, including surface width information, boundary type information, and boundary combination state. The surface width information characterizes the lateral or longitudinal dimensions of the surface to be cleaned; the boundary type information distinguishes between framed boundaries, frameless boundaries, virtual boundaries, or mixed boundaries; and the boundary combination state describes the relationship between the left and right boundaries, such as framed on both sides, framed on one side, or frameless on both sides. Compared to existing technologies that only distinguish between framed and frameless windows with coarse-grained modeling, this application achieves refined identification and quantitative characterization of window features. Furthermore, based on the acquired surface features, a matching path planning strategy is selected. By recognizing multi-dimensional features such as surface width, boundary type, and boundary combination states, this technology can automatically match appropriate path planning strategies for different types of surfaces to be cleaned (e.g., narrow windows, wide windows, framed windows, frameless windows, mixed boundary windows, etc.), significantly improving the adaptability of the window cleaning robot under complex surface conditions and overcoming the path pattern mismatch problem caused by existing technologies that only distinguish between framed and frameless windows. Furthermore, based on the selected path planning strategy, the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned. Since the selected path planning strategy is closely related to the characteristics of the surface to be cleaned, it can effectively avoid problems such as repeated wiping and partial omissions caused by path pattern mismatch with window type, ensuring full coverage cleaning of the surface and improving cleaning quality. Moreover, this application can adapt to scenarios with inconsistent boundary shapes, such as one side being framed and the other side being frameless, or situations with mixed boundaries in local areas. By dynamically adjusting the path planning strategy, it ensures stable execution of cleaning tasks under different boundary combination states, enhancing the robustness and reliability of the window cleaning robot in practical use.

[0187] Therefore, this application reduces ineffective turning, redundant paths, and repetitive cleaning behaviors of window cleaning robots by refining surface feature recognition and dynamic path selection. This is particularly effective in scenarios with narrow windows or asymmetrical boundary conditions, significantly shortening cleaning time and improving overall operational efficiency. Furthermore, the process does not rely on preset templates or manual intervention, enhancing the window cleaning robot's adaptability to different types of surfaces.

[0188] For example, Figure 1 This is a schematic diagram of an application scenario provided in an embodiment of this application, such as... Figure 1 As shown, the application scenario includes a user and a window cleaning robot 100. Taking the user starting the window cleaning robot 100 to clean the window glass as an example, at this time, the window cleaning robot 100 is attached to the window glass.

[0189] Furthermore, the window cleaning robot 100 can acquire the surface features of the window glass. For example, it can measure or calculate the lateral dimensions of the window surface through sensors, identify the window edge as a framed boundary, and determine that the combination of the left and right boundaries is a double-sided framed structure.

[0190] Furthermore, based on the surface features acquired above, the window cleaning robot 100 selects a strategy that matches the current surface features from a variety of preset path planning strategies. For example, a standard zigzag path can be selected. In this way, the window cleaning robot 100 performs cleaning operations on the window glass according to the selected zigzag path, including moving along the zigzag path, wiping, turning, and edge treatment, until full coverage cleaning is completed or the user interrupts the task.

[0191] In this way, by acquiring the window width, boundary type, and the boundary combination state with frames on both sides, the window cleaning robot 100 can adaptively select a matching Z-shaped path, thereby avoiding path misalignment, repeated wiping, or edge omissions caused by mismatched window features, and improving the integrity of cleaning coverage and operating efficiency.

[0192] It should be noted that the embodiments of this application do not limit the specific content of the surface features (such as surface width information, boundary type information, and boundary combination state) or the specific implementation method of the selected path planning strategy. The above example is only one possible implementation method. In practical applications, the type of surface features, the acquisition method, and the selection rules of the path planning strategy can all be adjusted or extended according to the specific scenario. Any equivalent substitution or variation based on the concept of this application falls within the protection scope of this application.

[0193] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0194] For example, Figure 2 This is a flowchart illustrating a control method for a window cleaning robot provided in an embodiment of this application, as shown below. Figure 2 As shown, the control method of this window cleaning robot includes the following steps:

[0195] S201. Obtain the surface features of the surface to be cleaned by the window cleaning robot. The surface features include surface width information, boundary type information, and boundary combination state.

[0196] In this embodiment, the surface to be cleaned can refer to the physical surface on which the window cleaning robot is currently adsorbed and ready to perform the cleaning operation. It can refer to the glass window surface of a building, but may also include other smooth surfaces (such as mirrors, tiled walls, etc.). This embodiment does not limit the specific surface type corresponding to the surface to be cleaned.

[0197] Surface features refer to a set of multidimensional parameters used to describe the geometry and boundary constraints of the surface to be cleaned. Surface features are a quantitative abstraction of the physical properties of the surface to be cleaned, including at least three dimensions: surface width information, boundary type information, and boundary combination state.

[0198] The surface width information refers to the effective cleanable distance of the surface to be cleaned in the robot's travel direction (usually left and right). For example, the surface width information may include the horizontal width and vertical distance information of the current window.

[0199] Optionally, the surface width information can be calculated using odometer data or distance sensor data as the window cleaning robot moves along the boundary. This application does not specifically limit the method of obtaining the surface width information.

[0200] Optionally, the surface width information can also be used to characterize the relationship between the current width of the surface to be cleaned and the body dimensions of the window cleaning robot.

[0201] It should be noted that surface width information can be used to determine whether a window is a "wide window" or a "narrow window", thus affecting the choice of path mode. For example, a narrow window may use a one-way round trip mode, while a wide window may use a standard Z-shaped path mode.

[0202] Boundary type information can refer to the classification identifier of the physical properties of the edge of the surface to be cleaned. Boundary type information can be used to distinguish whether the edge of the surface to be cleaned is a "framed boundary," a "frameless boundary," or a "virtual boundary." Optionally, boundary type information can be obtained through a collision sensor (detecting physical contact) or an edge detection sensor (detecting suspension). This application embodiment does not specifically limit the method of obtaining boundary type information.

[0203] Among them, framed boundaries refer to the physical borders (such as window frames or window sills) on the edges of the surface to be cleaned, allowing the window cleaning robot to safely contact and rely on the borders for turning or positioning.

[0204] Frameless boundaries refer to the edges of the surface to be cleaned being suspended (such as frameless glass). Window cleaning robots must avoid crossing these boundaries to prevent them from falling.

[0205] Virtual boundaries refer to non-physical boundary lines or area limits that are defined or identified by the window cleaning robot during the cleaning process, through sensor detection, algorithm calculation, or user settings. Virtual boundaries have no physical structure and do not depend on the physical edges of the surface being cleaned. They exist within the window cleaning robot's control algorithm and map data. For example, on frameless windows, the window cleaning robot uses drop sensors to detect the window edges and automatically generates a virtual boundary to prevent falls. Alternatively, the user can set a rectangular area through an application (APP), and the window cleaning robot will only clean within that rectangular area; the boundary of this area is the virtual boundary.

[0206] Understandably, virtual boundaries can be generated, adjusted, or deleted in real time based on the conditions of the surface to be cleaned, the requirements of the cleaning task, or user instructions.

[0207] Boundary combination state refers to the overall boundary shape description formed by combining the boundary type information of the left and right sides of the surface to be cleaned. Boundary combination state is a comprehensive description of the symmetry and consistency of the boundaries of the surface to be cleaned.

[0208] Optional, boundary combination states include:

[0209] Double-sided framed: Both the left and right boundaries are framed boundaries, and / or both the top and bottom boundaries are framed boundaries.

[0210] Double-sided frameless: Both the left and right boundaries are frameless, and / or both the top and bottom boundaries are frameless.

[0211] A border on one side and no border on the other: The border types of the left and right boundaries are inconsistent, and / or the border types of the top and bottom boundaries are inconsistent.

[0212] It should be noted that both framed and frameless also include corresponding virtual borders.

[0213] For example, after the window cleaning robot is placed and adsorbed onto the surface to be cleaned (such as a glass window), its sensors are activated. The control and processing unit of the window cleaning robot controls the robot to perform initial pose adjustment and establish an initial state machine corresponding to the current cleaning task. At the same time, it scans or detects the surface it is currently on using its onboard sensors (such as edge detection sensors, collision sensors, laser rangefinders, attitude sensors, or optical sensors). The control and processing unit then calculates and analyzes the collected sensor data, extracting three surface feature information related to the window surface, including:

[0214] Surface width information: The effective cleaning width of the surface to be cleaned is calculated by measuring the movable range or boundary distance on the left and right sides of the window cleaning robot.

[0215] Boundary type information: Determine whether the boundary that the window cleaning robot comes into contact with or detects is a physical border (with a frame) or a suspended edge (without a frame).

[0216] Boundary combination state: Combining the boundary type information on the left and right sides, a combination state describing the shape of the entire window boundary is formed (such as "left frame right frame", "left frame right none", "left none right none", etc.).

[0217] Optionally, in addition to extracting the three surface feature information mentioned above, the window cleaning robot can also obtain the current starting point position, current position, and relative boundary distance during the initial operation phase, as well as the current cleaning mode, task mode, and whether there is historical breakpoint information. Among them, the relative boundary distance is used to support the window cleaning robot in selecting strategies such as path switching, breakpoint recovery, and returning to the starting point.

[0218] It should be noted that, in addition to the automatic identification and processing by the window cleaning robot, the acquisition of the above-mentioned surface feature information can also be achieved by allowing users to select surface features through the terminal device's APP; or the surface features can be processed by a cloud model and the processing results can be sent to the window cleaning robot. The window cleaning robot only executes the path planning strategy and is not responsible for the judgment of surface features and the determination of the strategy.

[0219] S202. Select the corresponding path planning strategy according to the surface characteristics, and control the window cleaning robot to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

[0220] In this embodiment of the application, the path planning strategy may refer to a set of predefined, executable motion path patterns and their control rules adopted by the window cleaning robot in order to complete the task of full coverage cleaning of the surface to be cleaned.

[0221] It should be noted that the path planning strategy specifies how the window cleaning robot moves on the surface to be cleaned, when to turn, how to cover boundary areas, and how to return to the starting point. Different path planning strategies are suitable for different window characteristics, and their core objective is to achieve a balance between the integrity and efficiency of cleaning coverage while ensuring safety.

[0222] Optionally, the path planning strategy may include Z-shaped path strategy, N-shaped path strategy, one-way round-trip path strategy, and local partition cleaning path strategy, etc. The specific strategy corresponding to the path planning strategy is not limited in the embodiments of this application, and can be referred to the description in the following embodiments.

[0223] Among them, the Z-shaped path strategy is for the window cleaning robot to move back and forth in a zigzag pattern, using the side edges for efficient turning, covering a large area and with high efficiency.

[0224] The N-shaped path strategy involves the window cleaning robot first moving downwards longitudinally, then moving laterally, and then moving upwards longitudinally, repeating this process to form a continuous "N"-shaped coverage path.

[0225] The one-way round-trip path strategy involves the window cleaning robot moving in a single direction and returning along the same path after reaching the boundary. This avoids making large turns at the frameless edge, prioritizing safety and preventing falls.

[0226] The local partition cleaning path strategy divides the window into multiple independent regions and applies the above-mentioned adapted path strategy to each region to ensure that each partition is completely covered.

[0227] For example, after acquiring surface features, the window cleaning robot's control processing unit matches the received surface features with a preset "surface feature-strategy mapping rule base." This rule base stores the mapping relationship between different combinations of surface features and corresponding path planning strategies. Then, based on the matching result, a suitable path planning strategy for the current surface to be cleaned is selected from the rule base. For example, if the surface features are "wide window and framed on both sides," a "Z-shaped path strategy" is selected. If the surface features are "narrow window and framed on both sides," an "N-shaped path strategy" is selected. Further, based on the selected path planning strategy, the window cleaning robot generates specific motion control commands (such as movement direction, turning timing, and travel speed) and drives the window cleaning robot to perform the corresponding cleaning operation on the surface to be cleaned.

[0228] In contrast to existing technologies that only distinguish between "framed windows" and "frameless windows," this approach can easily lead to path pattern mismatches, repeated wiping, partial omissions, or decreased efficiency when dealing with narrow windows, narrow windows with frames on both sides, or windows with inconsistent boundary shapes. This application introduces a more refined surface feature recognition mechanism, recognizing not only boundary type information (framed / frameless) but also surface width information and boundary combination states. By recognizing surface width information, the window cleaning robot can distinguish between different types of surfaces to be cleaned, such as wide windows and narrow windows, thus identifying window types that existing technologies cannot distinguish, such as narrow windows and narrow windows with frames on both sides. Furthermore, by recognizing boundary combination states, the window cleaning robot can identify surfaces with inconsistent boundary shapes. Therefore, through these two added dimensions, the window cleaning robot can automatically adapt to diverse surfaces to be cleaned, significantly expanding its applicable scenarios and improving its cleaning adaptability to complex surfaces.

[0229] Furthermore, this application dynamically selects the corresponding path planning strategy based on the acquired surface features, rather than executing a fixed template path. This decision-making method ensures a precise match between the path pattern and the surface features to be cleaned, effectively eliminating edge omissions and uncovered areas caused by mismatches between the path pattern and the type of surface to be cleaned, thus improving the integrity of cleaning coverage. It also avoids ineffective turns and repeated wiping caused by fixed template paths in narrow window or asymmetrical boundary scenarios, shortening the overall cleaning time. The execution of this differentiated path strategy enables the window cleaning robot to achieve an effective balance between coverage integrity and cleaning efficiency for different types of surfaces to be cleaned.

[0230] Optionally, a corresponding path planning strategy can be selected based on surface features, including:

[0231] If the surface width information is determined to be greater than the preset width threshold, the path planning strategy is determined to be the first strategy, and the main travel direction of the cleaning path corresponding to the first strategy is horizontal.

[0232] If the surface width information is determined to be less than or equal to the preset width threshold, and the boundary combination state is that both the left and right boundaries have borders, the path planning strategy is determined to be the second strategy, and the main travel direction of the cleaning path corresponding to the second strategy is longitudinal.

[0233] In this embodiment, the preset width threshold can refer to a preset width threshold, which is related to the width of the window cleaning robot itself or the effective coverage width of the cleaning module. For example, it can be set to 1.5 times or 2 times the width of the window cleaning robot.

[0234] Optionally, a preset width threshold can be used to distinguish between "wide windows" and "narrow windows". For example, when the surface width information is greater than the preset width threshold, it is considered a "wide window" and is suitable for horizontal movement; when the surface width information is less than or equal to the preset width threshold, it is considered a "narrow window" and may require switching the movement direction, making it suitable for vertical movement.

[0235] Here, "lateral" refers to the direction of movement parallel to the horizontal direction of the window cleaning robot, i.e., the left-right direction. For example, with the lateral direction as the main direction of movement, the window cleaning robot moves back and forth along the width of the window (left-right). This direction is suitable for wider windows, maximizing the coverage area in a single pass.

[0236] The longitudinal direction can refer to the direction of movement parallel to the vertical direction of the window cleaning robot, i.e., the up-and-down direction. For example, with the longitudinal direction as the main travel direction, the window cleaning robot moves back and forth along the height of the window (up and down). This direction is suitable for narrow windows, reducing the number of lateral movements, decreasing the proportion of ineffective turns, and utilizing the height advantage of the window for effective coverage.

[0237] In this application, the first strategy may refer to a path planning strategy applicable to wide window scenarios. Under the first strategy, the window cleaning robot mainly moves back and forth in the left and right direction, gradually moving downwards or upwards to make full use of the horizontal space of the wide window and achieve efficient coverage.

[0238] The second strategy refers to a path planning strategy suitable for narrow windows with frames on both sides (both left and right boundaries exist). Under this strategy, the window cleaning robot mainly moves back and forth in the vertical direction, gradually advancing to the left or right. This strategy avoids frequent turning within the narrow horizontal space, thereby improving cleaning efficiency and safety in narrow window scenarios.

[0239] It should be noted that the first strategy corresponds to the first full-coverage path pattern, which can be a Z-shaped path or a horizontal strip path. The second strategy corresponds to the second full-coverage path pattern, which can be an N-shaped path or a vertical strip path.

[0240] For example, during the window cleaning robot's cleaning of window glass, the control processing unit can compare the acquired surface width information with a preset width threshold. If the surface width information is greater than the preset width threshold, it is determined to be a wide window scenario, and the path planning strategy is determined to be the first strategy. If the surface width information is less than or equal to the preset width threshold, and the boundary combination state is that both the left and right boundaries have borders (i.e., framed on both sides), it is determined to be a narrow window scenario with a frame, and the path planning strategy is determined to be the second strategy.

[0241] Thus, this application introduces a preset width threshold as a decision boundary point and performs secondary judgment based on the boundary combination state, enabling adaptive path direction guidance for different types of surfaces to be cleaned. For example, depending on whether the window is a "wide window" or a "narrow window," the main travel direction of the cleaning path (horizontal or vertical) is automatically switched to match the window geometry. In scenarios with narrow windows and frames on both sides, switching the main travel direction to vertical avoids frequent ineffective turns in the narrow horizontal space, thereby significantly improving cleaning efficiency. In wide window scenarios, using a horizontal main travel direction leverages the window's width to cover a larger area in a single pass, accelerating the overall cleaning speed and improving coverage speed in wide window scenarios.

[0242] Furthermore, this application enhances the targeting of path planning by using a dual judgment of preset width threshold and boundary combination state. For example, it can avoid misusing horizontal paths in narrow window scenarios or misusing vertical paths in wide window scenarios, ensuring that the path strategy is accurately matched with the window features.

[0243] Optionally, based on the selected path planning strategy, the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned, including:

[0244] When the first strategy is adopted, the window cleaning robot is controlled to move back and forth laterally from the starting position, and each time it reaches the first target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a lateral strip path covering the surface to be cleaned.

[0245] The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on a horizontal strip path.

[0246] In this embodiment of the application, the starting position may refer to the initial placement position when the window cleaning robot begins to perform the cleaning operation.

[0247] It's important to note that the starting position is typically where the user attaches the window cleaning robot to the window surface, or it can be the position the robot has reached after initialization or its return journey. In path planning, the starting position is the initial reference point for the cleaning path. The starting position can also be called the origin position.

[0248] Reciprocating motion can refer to the back-and-forth motion that a window cleaning robot alternates between two opposite directions.

[0249] In this step, back-and-forth movement can refer to reciprocating motion along the lateral (left-right) direction. The window cleaning robot first moves to one side to the boundary, then changes direction and moves to the other side, repeating this process to form a zigzag or Z-shaped trajectory.

[0250] The first target boundary can refer to the lateral boundary that the window cleaning robot needs to reach and trigger a turning action during the lateral back-and-forth movement.

[0251] Optionally, the first target boundary can refer to the left and right boundaries of the window. When the window cleaning robot moves to the first target boundary, a turning movement operation is triggered. The first target boundary can be a physical border (with a frame), a detected overhanging edge (without a frame), or a virtual border.

[0252] Turning and moving operations can refer to a sequence of actions performed by a window cleaning robot after reaching a boundary in order to change its direction of travel.

[0253] For example, a steering maneuver typically includes the following steps:

[0254] Stop the lateral movement in the current direction.

[0255] Move longitudinally (up and down) by a preset step distance (usually equal to or slightly less than the width of the cleaning module).

[0256] Change the direction of lateral movement (e.g., from left to right).

[0257] The above operations ensure that the window cleaning robot can smoothly transition from one horizontal strip to the next.

[0258] A lateral strip path can refer to a clean path pattern consisting of a series of parallel lateral movement trajectories.

[0259] Understandably, horizontal strip paths are a typical path pattern in wide-window scenarios. The window cleaning robot moves back and forth horizontally, and then steps longitudinally after reaching the boundary each time, forming a series of parallel and adjacent horizontal strip tracks on the window surface. These strip tracks collectively cover the entire surface to be cleaned, ensuring no area is missed.

[0260] For example, after the window cleaning robot begins its cleaning task from its initial placement position (starting point), it moves back and forth laterally (left and right). That is, it first moves to one side to the boundary, then turns to the other side to complete a cleaning strip, repeating this process. During each movement, when the window cleaning robot reaches the first target boundary, it performs a turning operation and path switching. Through this repeated movement, the window cleaning robot forms a series of parallel horizontal strip paths on the window surface, gradually covering the entire window surface. Simultaneously, while moving along the horizontal strip paths, the window cleaning robot activates cleaning modules (such as wiping cloths and squeegees) to perform cleaning operations on the surface to be cleaned.

[0261] Optionally, after cleaning is completed based on the lateral strip path, a return-to-start strategy matching the first strategy can be adopted. This is suitable for wide-window scenarios, reduces ineffective turning, and improves coverage efficiency per unit area.

[0262] Therefore, this application concretizes the first strategy into an execution process of lateral back-and-forth movement and boundary turning, achieving efficient coverage of the surface to be cleaned and ensuring the integrity of the coverage. For example, in wide-window scenarios, using lateral back-and-forth movement can maximize the coverage area of ​​a single trip and reduce the number of longitudinal movements, thereby significantly improving cleaning speed. Moreover, by performing a turning movement operation after each boundary is reached, it is ensured that there are no omissions between adjacent lateral strips, forming a continuous and complete coverage path.

[0263] Furthermore, by simplifying complex path planning into a loop of lateral movement and boundary turning, the implementation difficulty of the control logic is reduced, and the execution reliability is improved. This makes the first strategy particularly suitable for windows with large widths and regular boundaries, thus fully leveraging the efficiency advantages of lateral movement.

[0264] Optionally, the method also includes:

[0265] After the window cleaning robot finishes cleaning the surface to be cleaned based on the horizontal strip path, the robot is controlled to perform an edge cleaning strategy and return to the starting position.

[0266] In this embodiment, the edge cleaning strategy can refer to a supplementary cleaning path planning strategy that the window cleaning robot performs specifically for the surrounding boundary areas of the surface to be cleaned after the main body of the surface to be cleaned has been cleaned.

[0267] Optionally, the core objective of an edge cleaning strategy is to clean edge areas that are difficult to fully cover along the main path. It typically employs a boundary-following approach, allowing the window cleaning robot to move along the four edges or borders of the window and clean the areas near those edges. An edge cleaning strategy can perform one or more circles to ensure that edge areas are thoroughly cleaned.

[0268] For example, the execution flow corresponding to the edge cleaning strategy includes:

[0269] The window cleaning robot is controlled to adjust its body posture and move to detect the position and distance of the upper boundary, establishing a starting reference point. This confirms at least one followable boundary (such as a framed boundary or a detectable frameless edge). Once a followable boundary is confirmed, the robot can be controlled to adjust its posture to suit edge-following operation (e.g., parallel to the boundary). Furthermore, starting from the starting edge, it follows the top, right, bottom, and left edges in a clockwise direction, completing the cleaning of all four edges. Finally, after completing a preset number of circles (usually one), the edge-following cleaning ends, and the robot switches to a return-to-start strategy.

[0270] When transitioning from one edge to an adjacent edge, a specific turning action (such as rotating 90 degrees in place) can be performed to ensure a smooth transition.

[0271] In this way, the edge cleaning strategy can compensate for the insufficient coverage of the main path in the edge area, and targeted cleaning can be carried out on areas prone to dust accumulation such as borders and corners.

[0272] For example, after the window cleaning robot completes full-coverage cleaning of the surface to be cleaned based on a transverse strip path, it triggers the entry into the edge cleaning process. This involves executing an edge cleaning strategy, the corresponding process of which includes: the control processing unit performing initial posture adjustment and upper boundary detection, establishing the starting position, upper boundary distance, and boundary markers. After confirming the existence of at least one followable boundary, the robot body is adjusted to a posture suitable for edge-following operation, and it enters the edge-following process from the preset starting edge. Further, the window cleaning robot is controlled to perform edge cleaning, and after edge cleaning is completed, the robot is controlled to perform a return motion, returning to the starting position.

[0273] Thus, this application incorporates an edge cleaning strategy after the main body cleaning and returns to the starting position. The main body cleaning (e.g., a horizontal strip path) primarily covers the central area of ​​the window, while the edges of the surface to be cleaned (especially near the four edges) are prone to cleaning blind spots due to path turning or boundary detection errors. The edge cleaning strategy can specifically target these edge areas for secondary cleaning, effectively eliminating residual dirt. Therefore, by designing a two-stage cleaning mode of main body cleaning and edge cleaning, comprehensive coverage from the center to the edges is achieved, significantly improving the overall cleaning effect. Furthermore, by controlling the window cleaning robot to return to the starting position from a known boundary location after edge cleaning, the risk of falling or efficiency reduction due to path uncertainty during the return trip is reduced.

[0274] In addition, for framed windows, dust and stains tend to accumulate near the edges. Edge cleaning strategies can target the edges of the windows, meeting users' higher requirements for window edge cleanliness.

[0275] Optionally, based on the selected path planning strategy, the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned, including:

[0276] When the second strategy is adopted, the window cleaning robot is controlled to move back and forth in the longitudinal direction, and each time it reaches the second target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a longitudinal strip path covering the surface to be cleaned;

[0277] The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on a longitudinal strip path.

[0278] In this embodiment of the application, the second target boundary may refer to the longitudinal boundary that the window cleaning robot needs to reach and trigger a turning action during the longitudinal reciprocating movement.

[0279] Optionally, the second target boundary typically refers to the upper and lower boundaries of the window surface. When the window cleaning robot moves to the second target boundary, a turning movement operation is triggered. The second target boundary can be a physical border (with a frame) or a detected overhanging edge (without a frame), or it can be a virtual border. Corresponding to the first target boundary (horizontal boundary), the second target boundary is the turning trigger point for vertical movement.

[0280] In this step, the turning and movement operation refers to a sequence of actions performed by the window cleaning robot after reaching the boundary to change its direction of travel. In the second strategy, the specific content of the turning and movement operation is similar to that of the first strategy, but the direction is different. For example, the turning and movement operation may include:

[0281] Stop vertical movement in the current direction;

[0282] Move horizontally (left and right) by a preset step distance (usually equal to or slightly less than the width of the cleaning module).

[0283] Change the vertical movement direction (e.g., from upward to downward).

[0284] The above operations ensure that the window cleaning robot can smoothly transition from one longitudinal strip to the next.

[0285] A longitudinal strip path can refer to a clean path pattern consisting of a series of parallel longitudinal movement trajectories.

[0286] It's important to note that vertical strip paths are a typical path pattern for narrow window scenarios. The window cleaning robot moves back and forth along the longitudinal direction, and then steps laterally after reaching the upper and lower boundaries each time, forming a series of parallel and adjacent vertical strip-shaped tracks on the window surface. These strip-shaped tracks collectively cover the entire surface to be cleaned, ensuring no area is missed. Compared to lateral strip paths, vertical strip paths are more suitable for windows with limited width.

[0287] For example, after the window cleaning robot begins its cleaning task from its initial placement position (starting point), it moves back and forth longitudinally (up and down). That is, it first moves upwards or downwards to the boundary, then turns and moves in the opposite direction, repeating this process. During each movement, when the window cleaning robot reaches the second target boundary, it performs a turning movement operation. In this way, through the cycle of longitudinal movement, reaching the boundary, turning, and reverse longitudinal movement, the window cleaning robot forms a series of parallel longitudinal strip paths on the window surface, gradually covering the entire window surface. Simultaneously, as the window cleaning robot moves along the longitudinal strip paths, the cleaning module is activated to perform cleaning operations on the surface to be cleaned.

[0288] Optionally, a return-to-start strategy that matches the second strategy can be adopted, thereby adapting to the narrow window structure and reducing the risk of redundant paths and edge omissions caused by large-angle lateral switching.

[0289] Thus, this application improves the cleaning efficiency and coverage of the surface to be cleaned by specifying the second strategy as a longitudinal back-and-forth and boundary-turning execution process. It is particularly suitable for scenarios with narrow windows and frames on both sides. In this scenario, using longitudinal back-and-forth movement can avoid frequent and ineffective turns in the narrow lateral space, thereby significantly improving cleaning efficiency. Moreover, the longitudinal strip path can better conform to the geometry of the narrow window, ensuring no omissions in the coverage of the window surface with limited width. Compared to lateral back-and-forth, longitudinal back-and-forth can take advantage of the height of the window surface, allowing a longer distance to be covered in a single trip, reducing the number of turns, and reducing the time loss and path overlap caused by turns.

[0290] Furthermore, in the scenario described above where the window is narrow and has frames on both sides, the second strategy, by using longitudinal movement, allows the window cleaning robot to be closer to the frame on both sides, and the turning action is performed laterally, which helps to reduce the risk of falling due to excessive turning.

[0291] Optionally, before controlling the window cleaning robot to move back and forth longitudinally, the method further includes:

[0292] Control the window cleaning robot to perform a lateral edge probing and width measurement operation to obtain the available lateral width of the surface to be cleaned;

[0293] After obtaining the available horizontal width, the body posture of the window cleaning robot is adjusted;

[0294] If the robot body is found to be obstructed during the posture adjustment process, the robot body is controlled to retreat longitudinally by a preset safe distance, and the robot body posture is readjusted again so that the front side of the robot body contacts the boundary of the surface to be cleaned.

[0295] Based on the type of signal generated when the front side of the fuselage contacts the boundary, determine whether there is a restricted area on the surface to be cleaned;

[0296] After determining that there are restricted areas on the surface to be cleaned, the number of strips for the longitudinal strip path is determined based on the available lateral width.

[0297] In this embodiment of the application, the lateral edge probing and width measurement operation can refer to the operation in which the window cleaning robot moves laterally (left and right) before performing the main cleaning to detect the position of the left and right side boundaries, thereby measuring the available lateral width of the window surface.

[0298] It should be noted that the lateral edge detection and width measurement operation can be performed by moving the window cleaning robot to one side until the boundary detection sensor is triggered, recording the position of the boundary on that side, and then moving in the opposite direction to the other side boundary and recording the position of the other side boundary, thereby calculating the effective distance between the left and right boundaries, i.e. the lateral usable width.

[0299] The available lateral width refers to the effective distance that a window cleaning robot can safely move and clean in the lateral (left-right) direction on the surface to be cleaned.

[0300] Optionally, the available lateral width is the actual cleanable area after excluding borders, obstructions, or overhanging edges. It is a key parameter determining the number of stripes in the longitudinal strip path.

[0301] Steering obstruction refers to a situation where a window cleaning robot is unable to complete its intended turning action when attempting to adjust its body posture (such as rotation) due to insufficient space or interference with the boundary.

[0302] It should be noted that steering obstruction usually occurs in narrow windows or locally confined areas. When the length of the window cleaning robot is greater than the width of the window, it may be unable to complete a 90-degree or 180-degree rotation, resulting in posture adjustment failure.

[0303] The preset safety distance refers to a pre-set distance value used to reverse longitudinally when steering is obstructed. The preset safety distance can be set to half the length of the window cleaning robot's body or a fixed value (such as 10 centimeters).

[0304] In this step, the purpose of reversing is to create more space for the fuselage to rotate, avoiding collisions or falls caused by forced turning.

[0305] Signal type refers to the different types of trigger signals generated by sensors when the front side of the window cleaning robot contacts the boundary of the surface to be cleaned. Signal type is used to distinguish the nature of boundary contact and can include: front impact plate trigger signal and front boundary ball head disengagement signal.

[0306] The front impact trigger signal is triggered by the collision sensor on the front of the fuselage being triggered by the physical frame, indicating contact with a framed boundary. The front boundary ball joint detachment signal is triggered by the boundary detection sensor (such as a ball joint switch) on the front of the fuselage detecting suspension or edge detachment, indicating contact with a frameless boundary or corner area.

[0307] By analyzing the signal type, the window cleaning robot can determine whether the current scene is narrow-face restricted (with a frame) or corner restricted (without a frame or mixed boundaries).

[0308] A restricted area refers to a localized area on the surface to be cleaned that cannot be cleaned by the window cleaning robot following a standard path pattern due to insufficient width, complex boundary shapes, or the presence of obstacles. Restricted areas can include narrow window areas, corner areas, and areas with inconsistent boundary shapes.

[0309] Understandably, identifying restricted areas helps window cleaning robots adopt more conservative or adaptive path strategies, avoiding cleaning omissions or safety risks.

[0310] The number of strips refers to the total number of longitudinal movement trajectories required to cover the entire surface to be cleaned in longitudinal strip path planning. The number of strips is determined by the available lateral width and the robot's single cleaning coverage width. Determining the number of strips directly affects the cleaning coverage density and total time.

[0311] For example, after the window cleaning robot starts its cleaning task from the starting position, before beginning its longitudinal back-and-forth movement, it can first perform a lateral movement to detect the left and right boundaries and obtain the available lateral width of the surface to be cleaned. After obtaining the available lateral width, the window cleaning robot attempts to adjust its body posture (such as rotating to be parallel to the boundary) to prepare for subsequent longitudinal movement. Furthermore, if the robot's turning is obstructed during posture adjustment (such as the robot being too long or the space being too narrow to complete the rotation), the window cleaning robot retreats longitudinally by a preset safety distance to free up more space. After retreating, it attempts to adjust its body posture again so that the front of the robot contacts the boundary of the surface to be cleaned (such as the upper or lower boundary). Then, based on the signal type generated when the front of the robot contacts the boundary, it is determined whether there is a restricted area on the surface to be cleaned. If the generated signal type is a front collision trigger signal, it indicates that the front of the robot directly collides with the physical frame, which usually corresponds to narrow-face scenarios. If the generated signal type is a front boundary ball head detachment signal, it indicates that the front of the robot crosses the frameless edge, causing adsorption or contact failure, which usually corresponds to corner-restricted scenarios.

[0312] Furthermore, after determining that there are restricted areas on the surface to be cleaned, the number of strips for the longitudinal strip path is calculated and determined based on the obtained lateral available width to ensure complete coverage within the restricted space.

[0313] Thus, this application improves path planning accuracy and enhances the robustness of robot posture adjustment by adding lateral edge probing, attitude adjustment, and restricted area judgment before longitudinal cleaning to determine the number of strips in the longitudinal strip path. For example, by performing lateral edge probing and width measurement, the actual usable width of the window surface can be accurately obtained, providing accurate data for subsequent calculation of the number of longitudinal strips and avoiding coverage omissions or path overlaps caused by width estimation errors. By detecting steering obstruction and performing a reversal operation, the problem of the robot's inability to complete attitude adjustment in narrow spaces is solved, ensuring that the window cleaning robot can smoothly enter the longitudinal cleaning mode.

[0314] Furthermore, by analyzing the signal type when the fuselage's front side contacts the boundary, this application can distinguish between two different scenarios: narrow-area confinement and corner confinement, providing more refined environmental information for subsequent path planning. Based on the available lateral width and confined area information, the number of longitudinal stripes is dynamically determined, ensuring reasonable coverage density within a limited width. This avoids efficiency degradation due to too many stripes or omissions due to too few strips, thus optimizing the coverage strategy for longitudinal strip paths.

[0315] Optionally, a corresponding path planning strategy can be selected based on surface features, including:

[0316] Obtain the current task type of the window cleaning robot, and select the corresponding path planning strategy based on the current task type and surface features.

[0317] In this embodiment, the current task type refers to the category of cleaning task that the window cleaning robot is currently performing or about to perform. Different task types correspond to different path planning strategies and cleaning execution logic.

[0318] Optionally, the current task type may include at least global cleaning tasks, edge cleaning tasks, and local area cleaning tasks. This application embodiment does not limit the specific type corresponding to the current task type. For example, the current task type may also include fixed-point cleaning tasks and remote-controlled cleaning tasks. Fixed-point cleaning tasks are used to repeatedly wipe a specified local area. Remote-controlled cleaning tasks may refer to cleaning an area based on manual intervention control.

[0319] In this context, a global cleaning task can refer to a complete cleaning task designed to cover all usable areas of the surface to be cleaned, used to perform full-coverage cleaning of the entire surface.

[0320] Understandably, global cleaning is a common cleaning mode. When performing a global cleaning task, a window cleaning robot needs to plan a path that covers the entire window surface (from edge to edge, from top to bottom) to ensure no area is missed. For example, global cleaning tasks typically employ Z-shaped, N-shaped, or strip-shaped path strategies.

[0321] Edge cleaning refers to tasks specifically designed to clean the perimeter areas of a surface. It's used to clean edge areas that are difficult to reach via the main path. When performing edge cleaning tasks, window cleaning robots typically employ a boundary-following path strategy, moving along the four edges to provide supplementary cleaning.

[0322] Optionally, the edge cleaning task can be executed independently or automatically triggered after the global cleaning task is completed. This application does not specifically limit this.

[0323] Localized area cleaning tasks can refer to cleaning a specific area on a surface to be cleaned.

[0324] Optionally, the localized cleaning task allows the user to specify a localized area (e.g., by selecting a box via an app or by the robot automatically identifying stains). The window cleaning robot then cleans only that localized area, without covering the entire window surface. When performing localized cleaning tasks, the window cleaning robot typically employs a localized partitioning cleaning path strategy, such as a spiral path, a reciprocating path, or an adaptive path based on area boundaries. This application embodiment does not specifically limit this approach.

[0325] For example, after the window cleaning robot acquires surface features, before or simultaneously with selecting and executing the corresponding path planning strategy based on those features, the robot's control processing unit acquires its current task type. Further, the surface features and the current task type are treated as two independent decision dimensions, and an appropriate path planning strategy is selected based on a preset mapping rule base. For instance, if the current task type is a global cleaning task, and the surface features are a wide window with frames on both sides, a Z-shaped path strategy is selected. If the current task type is a local area cleaning task, and the surface features are a narrow window with frames on both sides, a localized cleaning path strategy is selected. The window cleaning robot then performs the cleaning operation based on the selected path planning strategy.

[0326] It should be noted that when the window cleaning robot detects that the left and right boundaries are not established simultaneously, the boundary information is insufficient, or there are obvious frameless edge signals, the window cleaning robot will execute different strategies according to the current task type and surface features.

[0327] In this way, by introducing the current task type as an additional decision-making dimension for path planning strategy selection, this application enables the window cleaning robot to adjust its path strategy according to the specific task to be completed, rather than using the same path pattern for all tasks, thus achieving task-oriented differentiated paths. For example, in the edge cleaning task, the window cleaning robot can directly enter the edge cleaning path without first executing the global lateral strip path, thereby avoiding unnecessary path waste and improving task execution efficiency.

[0328] Furthermore, when users select different task types based on their actual needs, the window cleaning robot can automatically adapt its path strategy according to the selected task type, thus improving the product's ease of use and intelligence.

[0329] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0330] If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that at least one of the left and right boundaries has a border, control the window cleaning robot to execute the edge cleaning strategy so that the window cleaning robot moves and cleans along the boundary of the surface to be cleaned.

[0331] If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that neither the left nor the right boundary has a border, or the surface width information is less than or equal to a preset threshold, the window cleaning robot is controlled to execute a frameless return-to-start strategy so that the window cleaning robot returns to the starting position.

[0332] In this embodiment, the preset threshold can refer to a predetermined width threshold used to determine whether the width of the surface to be cleaned is sufficient to support safe edge cleaning. The preset threshold is typically related to the width of the window cleaning robot's body or its minimum safe turning radius. This embodiment does not specifically limit the size of the preset threshold.

[0333] Optionally, when the surface width information is less than or equal to a preset threshold, it is considered an "extremely narrow window surface". In this case, the window cleaning robot cannot safely perform edge cleaning or boundary following operations on the window surface, so it is necessary to switch to the frameless return-to-start strategy.

[0334] The frameless return-to-start strategy refers to a path planning strategy in scenarios with frameless boundaries or extremely narrow windows, where the window cleaning robot abandons edge cleaning and instead performs a series of safe actions to return to the starting position.

[0335] It should be noted that the core purpose of the frameless return-to-start strategy is to ensure that the window cleaning robot can safely and controllably return to the starting point in the absence of reliable physical boundaries, avoiding falling or getting lost.

[0336] Optionally, the execution flow corresponding to the frameless return-to-start strategy includes:

[0337] Adjust the window cleaning robot to a safe orientation, avoiding pointing it towards the frameless edge.

[0338] Control the window cleaning robot to move vertically, detect the upper boundary position and record the reference point.

[0339] Based on the measured distance from the top, retreat longitudinally to a safe position.

[0340] Control the window cleaning robot to move laterally, detect the left and right boundaries, and perform lateral retraction.

[0341] Return the aircraft to its initial orientation or safe ending position to complete the return journey.

[0342] It should be noted that the frameless return-to-start strategy does not rely on the closed-loop edge path, but instead uses local detection information (such as distance from the top edge and collision stop signal) to gradually guide the window cleaning robot back to the vicinity of the safe starting point, which is suitable for complex scenarios with incomplete boundaries or insufficient width.

[0343] For example, if the current task type is confirmed to be an edge cleaning task, and the boundary combination state is that at least one of the left and right boundaries has a border, i.e., there is at least one valid boundary reference that can be established, then it is determined to be a framed or partially framed scenario. In this case, the window cleaning robot is controlled to execute the edge cleaning strategy, causing the window cleaning robot to move and clean along the boundary of the surface to be cleaned. If the boundary combination state is that neither the left nor the right boundary has a border, or the surface width information (or the effective lateral width) is less than or equal to a preset threshold (or the body size threshold), then it is determined to be a frameless or extremely narrow scenario. In this case, the window cleaning robot is controlled to execute the frameless return-to-start strategy, causing the window cleaning robot to return to the starting position without performing edge cleaning.

[0344] Thus, this application differentiates the edge cleaning task in framed and frameless scenarios by selecting different path planning strategies. This not only improves the targeting and safety of edge cleaning but also avoids ineffective edge following in frameless scenarios, optimizing task execution efficiency. Specifically, in scenarios with double-sided frameless or extremely narrow windows, executing an edge cleaning strategy carries the risk of falling or failing to form a closed-loop path. By switching to a frameless return-to-start strategy, the window cleaning robot avoids ineffective or dangerous boundary following at frameless edges. Furthermore, performing edge cleaning when there is at least one framed boundary ensures that the window cleaning robot has a reliable physical boundary to follow, thereby improving the safety and effectiveness of edge cleaning.

[0345] In addition, this application identifies extremely narrow window surfaces by setting a preset threshold, avoiding edge-to-edge operations on window surfaces that are not wide enough to support safe edge-to-edge movement, reducing the risk of falls or collisions, and improving the robustness of the window cleaning robot in performing tasks.

[0346] It should also be noted that in frameless or extremely narrow scenarios, controlling the window cleaning robot to return directly to the starting point instead of attempting to perform edge cleaning can also avoid long periods of ineffective movement or getting stuck due to the inability to complete the edge path, thus improving the overall task execution efficiency.

[0347] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0348] If the current task type is determined to be a global cleaning task, and the boundary combination state is determined to be that both the left and right boundaries have borders, and the surface width information is greater than a preset threshold, the window cleaning robot is controlled to execute a full-coverage path strategy.

[0349] If the current task type is determined to be a global cleaning task, and the surface features are determined to meet at least one of the following conditions: the boundary combination state is that at least one of the left and right boundaries has no border, the surface width information is less than or equal to a preset threshold, or no surface width information is detected, then the window cleaning robot is controlled to execute a protection strategy. The protection strategy includes at least one of the following: a backtracking strategy, a return to the starting point strategy, or termination of the cleaning operation.

[0350] A full-coverage path strategy refers to a path planning strategy designed to cover the entire usable area of ​​the surface to be cleaned. It is the standard path pattern for global cleaning tasks.

[0351] Optionally, a full-coverage path strategy can include a Z-shaped path, an N-shaped path, or a strip path, using back-and-forth movement and step-by-step progression to ensure that the entire window surface is covered without omission. It should be noted that the full-coverage path strategy is suitable for security windows that are wide enough, have regular boundaries, and are framed on both sides.

[0352] The retreat strategy can refer to a protective action in which a window cleaning robot moves a safe distance in the opposite direction of its current movement when it detects unsafe conditions, in order to move away from the danger zone.

[0353] It should be noted that when the window cleaning robot detects an unlined edge or insufficient width in front of it, it immediately stops its current movement and retreats in the opposite direction to a safe position. The retreat distance can be a preset fixed value or dynamically adjusted based on sensor data; this embodiment does not specifically limit this.

[0354] The "return to starting point" strategy can refer to a protective strategy in which a window cleaning robot abandons its current cleaning task and performs a series of safety actions to return to its starting position when it detects unsafe conditions.

[0355] For example, when window conditions do not meet safety cleaning requirements, the window cleaning robot will no longer attempt to clean, but will instead plan a safe path back to the starting point. This return-to-start strategy typically involves moving through framed boundaries or detected safe areas, avoiding frameless edges or extremely narrow zones.

[0356] Optionally, after the window cleaning robot returns to its starting position, it can wait for the user to reposition it or switch task modes.

[0357] Optionally, the window cleaning robot can also be controlled to return to the starting position by external manual recall, remote control recall, or visual navigation retrieval.

[0358] For example, if the window cleaning robot confirms that the current task type is a global cleaning task, and the boundary combination state is that both the left and right boundaries have borders, and the surface width information is greater than a preset threshold, then it is determined to be a safe wide window scenario. In this case, with complete boundary and effective width information established, the window cleaning robot is controlled to execute a full-coverage path strategy to cover the entire surface to be cleaned. If the surface features meet at least one of the following conditions: the boundary combination state is that at least one of the left and right boundaries has no border (insufficient boundary constraints), the surface width information is less than or equal to a preset threshold, or no surface width information is detected, then it is determined to be an unsafe or abnormal scenario. In this case, the window cleaning robot is controlled to execute a protection strategy.

[0359] Optionally, if the window cleaning robot still cannot form a complete coverage geometry after the posture correction is determined, or if obvious frameless edges or cliff triggers occur during operation, the window cleaning robot can also be controlled to execute a protection strategy.

[0360] Therefore, if a window cleaning robot performs a global cleaning task on a frameless, extremely narrow, or surface with incomplete width information, there is a very high risk of it falling. Thus, by assessing the safety of the surface to be cleaned during a global cleaning task and implementing protective strategies for unsafe scenarios, the window cleaning robot effectively avoids accidents and prevents global cleaning from being performed on dangerous surfaces. Furthermore, by proactively identifying unsafe surfaces and taking protective measures, the risk of the window cleaning robot falling and being damaged is reduced.

[0361] Moreover, the protection strategy provides multiple options (reverse, return to starting point, end operation), allowing the window cleaning robot to choose a safe response method according to the specific scenario, thus enhancing the safety redundancy and robustness of the window cleaning robot.

[0362] It should also be noted that when surface features do not meet global cleaning requirements (such as extremely narrow windows), forcibly implementing a full-coverage path strategy may lead to path overlap, inefficiency, or omissions in coverage. This application addresses the situation where surface features do not meet global cleaning requirements by implementing a protection strategy to avoid invalid cleaning operations.

[0363] Optionally, the current path planning can be terminated if the complete coverage condition is not met. The failure to meet the complete coverage condition means that the window cleaning robot cannot simultaneously obtain the complete boundary constraints and effective geometric quantities required to execute the full coverage path strategy in the current scenario, or that obvious frameless edges, cliff risks, persistent boundary loss, or posture mismatch occur during path execution.

[0364] Specifically, it may include at least one of the following situations:

[0365] The left and right boundaries were not established simultaneously, making it impossible to form a stable lateral constraint;

[0366] The effective lateral width is less than the fuselage size threshold, which cannot support the implementation of a full-coverage path strategy;

[0367] Key geometric quantities (such as the lateral width of the surface to be cleaned, the top distance, and the left distance) are missing or unreliable;

[0368] The window cleaning robot experienced persistent frameless edge signals, cliff triggering, or posture correction failure during path switching or operation.

[0369] The current regional target boundary has been reached, but continuing to implement the full coverage path strategy will introduce the risk of exceeding the boundary or redundant paths.

[0370] In the above situation, the current path planning can be terminated. The purpose is to avoid problems such as falling, going out of bounds, missing edges, or invalid repeated movements caused by continuing to execute the full coverage path strategy under insufficient constraints. It can also provide a deterministic exit for returning to the starting point, mode switching, or safe shutdown.

[0371] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0372] If the current task type is determined to be a local cleaning task, and the surface width information of the local area is determined to meet the preset conditions, the window cleaning robot is controlled to execute the first strategy.

[0373] If the current task type is determined to be a local cleaning task, and the surface width information of the local area does not meet the preset conditions, or the boundary combination state of the local area is that at least one of the left and right boundaries does not have a border, the window cleaning robot is controlled to execute a local strip coverage strategy so that the window cleaning robot can move back and forth to clean the local area based on the strip coverage path.

[0374] The preset conditions can refer to pre-defined width conditions used to determine whether the width of a local area is sufficient to support the first strategy (lateral back-and-forth movement). Optionally, the preset conditions can be that the surface width information is greater than a preset width threshold. This preset width threshold is related to the width of the window cleaning robot's body, the minimum safe turning radius, or the effective coverage width of the cleaning module.

[0375] Understandably, when the width of a local area meets the preset conditions, it is considered a wide window local area, and in this case, a horizontal back-and-forth path is suitable; when the width of a local area does not meet the preset conditions, it is considered a narrow window local area, and other path modes can be used.

[0376] The local strip coverage strategy refers to a path planning strategy in local cleaning tasks that uses strip coverage paths for back-and-forth cleaning of local areas such as narrow windows, frameless areas, or areas with incomplete boundaries.

[0377] Optionally, local stripe coverage strategies may include:

[0378] Based on the geometry of the local area, the main direction of the strip's travel (horizontal or vertical) is automatically selected to maximize coverage efficiency;

[0379] The step distance between strips is dynamically adjusted according to the width of the local area to ensure complete coverage without overlap.

[0380] In frameless or incomplete boundary areas, the strip path will reserve a safe distance to prevent the robot from approaching the frameless edge.

[0381] For example, a local strip coverage strategy could involve controlling the window cleaning robot to perform small-interval back-and-forth movements between adjacent strips, with the main direction of the local area as the longitudinal axis. After reaching the boundary of the local area each time, the window cleaning robot would reverse direction and switch paths on a small scale to enter the next strip; and the comparison between the width of the completed strip and the distance to the target area would serve as the termination condition.

[0382] Thus, compared to conventional global coverage strategies, local strip coverage strategies do not require the establishment of complete window borders. Instead, they rely on locally measurable distances and local boundary signals to complete coverage and cleaning of restricted areas within a smaller range.

[0383] Optionally, the local strip coverage strategy includes at least one of a local N-shaped path or a longitudinal reciprocating strip path, so that the window cleaning robot can perform reciprocating cleaning based on the strip coverage path in a local area.

[0384] For example, when the window cleaning robot confirms that the current task type is a local cleaning task, it can acquire the surface features of the local area specified by the user or identified by the window cleaning robot. The surface features include surface width information and boundary combination state. If the surface width information of the local area meets the preset conditions (such as being greater than a preset width threshold), it is determined to be a local area of ​​a wide window with frames on both sides. At this time, the window cleaning robot is controlled to execute the first strategy, moving back and forth in the local area for cleaning.

[0385] If the surface width information of a local area does not meet the preset conditions (such as being less than or equal to the preset width threshold), or if the boundary combination state of the local area is that at least one of the left and right boundaries does not have a border, it is determined to be a local area with a narrow window, no frame, or incomplete boundary. At this time, the window cleaning robot is controlled to execute a local strip coverage strategy, so that the window cleaning robot can move back and forth to clean within the local area based on the strip coverage path.

[0386] Among them, the local strip coverage strategy can correspond to an N-type path or a longitudinal round-trip strip path.

[0387] It should be noted that if the window cleaning robot determines that the current scene is more suitable for longitudinal back-and-forth movement, a local strip coverage strategy can also be adopted.

[0388] In this way, by selecting different path strategies based on the width and boundary conditions of the local area during local cleaning tasks, fine-grained path adaptation for local areas is achieved, avoiding the efficiency reduction or safety risks caused by misusing lateral paths in narrow windows or frameless local areas. Furthermore, for narrow windows or frameless local areas, the local strip coverage strategy ensures complete coverage without omissions, improving the overall coverage integrity of local cleaning. For wide windows with frames on both sides, the first strategy is used to fully utilize the width advantage, quickly completing the cleaning of the local area and optimizing the efficiency of local cleaning.

[0389] In addition, for frameless or incomplete boundary areas, a local strip coverage strategy can be adopted to prevent the magnetic storage robot from approaching the frameless edge due to lateral movement, reduce the risk of falling, and enhance the safety of local cleaning.

[0390] Optionally, a corresponding path planning strategy can be selected based on the current task type and surface features, including:

[0391] If the current task type is determined to be a zone cleaning task, the surface to be cleaned is divided into multiple target areas;

[0392] Select the corresponding path planning strategy based on the surface width and boundary type information of each target area.

[0393] In this embodiment, a zone cleaning task can refer to a task type that divides the surface to be cleaned into multiple independent target areas and performs cleaning operations on each area sequentially. Optionally, a zone cleaning task can also be used to clean a preset local area.

[0394] It should be noted that the zoned cleaning task is suitable for window surfaces with complex shapes, large width variations, or inconsistent boundary types. By dividing the surface to be cleaned into multiple regular sub-regions, the difficulty of path planning for individual regions can be reduced, and an appropriate path strategy can be selected for each region.

[0395] Optionally, performing a partition cleaning task may include steps such as partitioning, selecting partition strategies, and connecting paths between partitions.

[0396] A target area refers to an independent sub-area to be cleaned within a zonal cleaning task, defined according to preset rules. The target area is the basic execution unit of a zonal cleaning task. Each target area has independent surface characteristics (such as surface width information and boundary type) and corresponds to an independent path planning strategy.

[0397] Optionally, the rules for dividing the target region may include:

[0398] Classify according to the geometry of the surface to be cleaned (e.g., rectangular, L-shaped, T-shaped);

[0399] Divide according to boundary type (e.g., framed / frameless) or boundary combination state;

[0400] The area is divided according to the region specified by the user via the app or voice command;

[0401] The cleaning process is divided according to the location of obstacles (such as window handles and hinges) on the surface to be cleaned.

[0402] It should be noted that the embodiments of this application do not specifically limit the rules for dividing the target area; the above are merely illustrative examples.

[0403] For example, when the window cleaning robot confirms that the current task type is a zone cleaning task, the surface to be cleaned can be divided into multiple target areas according to preset rules. For instance, based on the set area orientation, the target window surface can be divided into any one of the following target areas: upper half, lower half, left half, or right half. For each target area, its corresponding surface width information and boundary type information are obtained. Then, based on the surface width information and boundary type information of each target area, the corresponding path planning strategy is selected. For example, for areas with a large width and frames on both sides, the first strategy is selected. For areas with a small width and frames on both sides, the second strategy is selected. For frameless areas or areas with incomplete boundaries, a protection strategy or a local strip coverage strategy is selected.

[0404] Furthermore, the window cleaning robot is controlled to execute the selected path planning strategy in each target area in the order of the divided areas to complete the cleaning of the entire surface.

[0405] It should be noted that for the partition cleaning task, this application adopts the same discrimination mechanism as the global cleaning task. However, when switching paths, it is necessary to additionally consider the area boundary, area direction and current area width of the local area, so as to adaptively select the Z-shaped, N-shaped or frameless strip path mode within the target area, and execute the edge-filling cleaning strategy or the return-to-starting-point strategy after the area coverage is completed.

[0406] Thus, this application achieves refined path planning for complex surfaces by dividing the surface to be cleaned into multiple target areas and adapting each area to a specific path planning strategy. This is particularly beneficial for complex window surfaces with irregular shapes, varying widths, or inconsistent boundary types, where a single path strategy cannot cover all areas. The zoned cleaning task allows for the selection of an appropriate path planning strategy for each area, enabling refined path planning.

[0407] Furthermore, by dividing the surface to be cleaned into multiple target areas and implementing corresponding path planning strategies for each, overall cleaning efficiency can be improved. For example, an efficient lateral back-and-forth strategy can be used in wide window areas, while a vertical strip strategy can be used in narrow window areas, avoiding efficiency losses caused by using a single strategy globally. For irregularly shaped windows such as L-shaped, T-shaped, and curved surfaces, the zoned cleaning task can divide the surface into multiple regular sub-areas, with each sub-area employing a corresponding path planning strategy, thereby achieving effective coverage of irregularly shaped windows.

[0408] In addition, by rationally dividing areas and connecting paths between areas, a smooth transition of cleaning paths between different areas can be ensured, reducing duplicate coverage or omissions, thereby optimizing the continuity of cleaning paths.

[0409] Optionally, the method also includes:

[0410] After the window cleaning robot has completed any of the tasks of zone cleaning, global cleaning, or edge cleaning, obtain the cleaning mode and boundary type information of the surface to be cleaned.

[0411] Based on cleaning mode and boundary type information, determine whether the window cleaning robot should perform an edge supplementary cleaning strategy.

[0412] In this embodiment of the application, the cleaning mode may refer to a preset working mode that the user can select to adjust the cleaning intensity and time consumption when the window cleaning robot performs cleaning tasks.

[0413] It's important to note that the cleaning mode is a crucial parameter for users interacting with the window cleaning robot, and it can typically be selected via an app, remote control, or buttons on the robot itself. Different cleaning modes correspond to different path planning strategies, cleaning cycles, and cleaning durations.

[0414] Optional cleaning modes include deep cleaning mode and quick cleaning mode. Deep cleaning mode aims to maximize cleaning effectiveness, performing a more comprehensive and thorough cleaning operation. Deep cleaning mode is suitable for scenarios where the surface to be cleaned is heavily soiled, has not been cleaned for a long time, or where users have high requirements for cleaning quality.

[0415] In deep cleaning mode, the window cleaning robot can perform at least one of the following operations:

[0416] Perform denser path coverage (e.g., reduce strip step distance);

[0417] After the main cleaning is completed, additional cleaning along the edges will be performed automatically;

[0418] This may increase the frequency of cleaning (such as performing secondary cleaning on the same area);

[0419] Slow down the cleaning speed to improve wiping effect.

[0420] Quick Clean mode is a working mode that aims to minimize cleaning time and perform streamlined and efficient cleaning operations. Quick Clean mode is suitable for scenarios where the surface to be cleaned is relatively clean, the user is short on time, and needs to complete the cleaning quickly.

[0421] In quick cleaning mode, the window cleaning robot can perform at least one of the following operations:

[0422] Standard path coverage is used, and no additional cleaning is required;

[0423] Skip to edge cleaning;

[0424] Improve cleaning speed;

[0425] Reduce the number of cleaning sessions (e.g., perform a coverage only once).

[0426] The supplementary edge cleaning strategy refers to a supplementary edge cleaning operation performed on the perimeter area of ​​the surface to be cleaned after the main cleaning task has been completed.

[0427] It should be noted that the edge-scraping cleaning strategy is an optional step in the deep cleaning mode. It is typically executed after the main cleaning (such as horizontal or vertical strip paths) is completed, to clean edge areas that are difficult to fully cover by the main path. Whether or not the edge-scraping cleaning strategy is executed depends on a comprehensive judgment of the cleaning mode and boundary type information.

[0428] For example, after the window cleaning robot completes any of the following tasks—zonal cleaning, global cleaning, or edge cleaning—it can obtain the cleaning mode (e.g., deep cleaning mode or fast cleaning mode) and boundary type information (e.g., framed, frameless, or combined boundary status) of the surface to be cleaned. Further, based on the cleaning mode and boundary type information, it determines whether to execute an edge supplementary cleaning strategy. For instance, if the cleaning mode is determined to be deep cleaning mode and the boundary type information is framed, edge supplementary cleaning is executed to thoroughly clean the edge of the frame. If the cleaning mode is determined to be fast cleaning mode and the boundary type information is framed, edge supplementary cleaning is not executed to save time. If the cleaning mode is determined to be deep cleaning mode and the boundary type information is frameless, edge supplementary cleaning is not executed to avoid the risk of falling. If the cleaning mode is determined to be fast cleaning mode and the boundary type information is frameless, edge supplementary cleaning is not executed.

[0429] Thus, this application determines whether the window cleaning robot should execute the edge-supplementary cleaning strategy by introducing dual judgment of cleaning mode and boundary type information after the completion of any of the partition cleaning task, global cleaning task, or edge-supplementary cleaning task. This avoids ineffective edge-supplementary cleaning in frameless scenarios and optimizes the total task time. For example, in frameless boundary scenarios, even in deep cleaning mode, the edge-supplementary cleaning strategy is skipped, preventing the window cleaning robot from performing dangerous or ineffective boundary following at frameless edges. In fast cleaning mode, skipping the edge-supplementary cleaning strategy can significantly shorten the total task time, meeting users' needs for rapid cleaning.

[0430] Moreover, based on the above control logic, users do not need to manually select whether to execute the edge supplement cleaning strategy. Instead, the window cleaning robot automatically makes decisions based on the cleaning mode and boundary type, which simplifies user operation and improves the flexibility of user experience.

[0431] Furthermore, after users select a cleaning mode according to their actual needs, the window cleaning robot can adjust the cleaning quality as required. Specifically, in deep cleaning mode, it automatically performs a supplementary edge cleaning strategy to ensure no blind spots at the edges; in quick cleaning mode, it skips the edge cleaning strategy to improve cleaning efficiency.

[0432] Optionally, after the window cleaning robot completes its main cleaning task, the robot's control processing unit can comprehensively determine whether to execute a supplementary edge cleaning strategy based on the current task type, main path type, cleaning mode, boundary status, and edge loop configuration. The specific judgment logic is as follows:

[0433] If the main cleaning task is a global cleaning task, and the main path is a Z-shaped path suitable for supplementary cleaning, and the number of edge loops configured in the current cleaning mode is greater than 0, then the edge supplementary cleaning strategy is executed; if the main path is an N-shaped path, then after the main cleaning is completed, the strategy of returning to the starting point can be executed directly.

[0434] For a main cleaning task that is a zone cleaning task, if the main path of the target area is a Z-shaped path, the number of edge loops configured in the current cleaning mode is greater than 0, and the required boundary of the target area exists, then the edge supplementary cleaning strategy is executed; otherwise, the return-to-start strategy is executed directly.

[0435] For the main cleaning task being an independent edge cleaning task, if no effective frame is detected or the window width is insufficient, the window cleaning robot will not perform the regular edge cleaning, but will instead perform a frameless return-to-start strategy.

[0436] Optionally, the method also includes:

[0437] If it is determined that the window cleaning robot will perform a supplementary cleaning strategy along the edges, control the window cleaning robot to perform the following operations:

[0438] Determine the starting boundary for edge-to-edge cleaning based on boundary type information;

[0439] The window cleaning robot is controlled to move sequentially along the starting boundary in a first preset direction to perform supplementary cleaning on the edge areas of the surface to be cleaned;

[0440] Record the number of times the window cleaning robot cleans along the edges, and when the number of times it cleans along the edges reaches the preset number, control the window cleaning robot to execute the return-to-start strategy.

[0441] In this embodiment, the starting boundary refers to the boundary that the window cleaning robot first follows and cleans when it begins supplementary edge cleaning. The selection of the starting boundary is usually based on boundary type information. For example, a framed boundary can be selected as the starting boundary to ensure that the window cleaning robot has a reliable physical boundary to follow.

[0442] Optionally, in frameless scenarios, the detected upper or lower boundary can be selected as the starting boundary. It should be noted that the choice of the starting boundary may affect the path order and safety of subsequent edge cleaning.

[0443] It should be noted that for the subsequent edge-supplement cleaning of the global cleaning task, the left boundary is used as the starting boundary by default; for the independent edge-supplement cleaning task, the right boundary is used as the starting boundary by default; and for the edge-supplement cleaning of the partition cleaning task, the left boundary is used as the starting boundary according to the fixed edge order.

[0444] The first preset direction can refer to the preset direction in which the window cleaning robot moves along the boundary when it performs supplementary cleaning. The first preset direction can be clockwise or counterclockwise. This application embodiment does not limit the specific direction corresponding to the first preset direction.

[0445] Optionally, the choice between clockwise or counterclockwise direction depends on the preset or window boundary characteristics. For example, for framed windows, a clockwise direction is typically chosen to ensure the window cleaning robot always moves along the inside of the frame, avoiding the risk of falling. The selection of the first preset direction should ensure the continuity and safety of the cleaning path along the edge.

[0446] The edge area can refer to the narrow strip-shaped area near the perimeter of the surface to be cleaned.

[0447] It should be noted that edge areas are areas that are difficult to completely cover by the main cleaning path (such as a horizontal or vertical strip path). Because the main path needs to turn or step at the boundary, edge areas are prone to cleaning blind spots. The edge supplementary cleaning strategy is specifically designed to clean edge areas, ensuring that there are no dead corners on the edges of the surface to be cleaned.

[0448] The preset number of edge cleaning cycles refers to the preset number of cycles required for edge cleaning supplementation.

[0449] It should be noted that the preset number of edge cleaning cycles can be 1 or 2 cycles. One cycle of edge cleaning can cover the basic cleaning needs of the edge area; two cycles of edge cleaning can further enhance the edge cleaning effect and are suitable for deep cleaning mode. The selection of the preset number of edge cleaning cycles should balance the cleaning effect and the task time. Therefore, the specific number of cycles corresponding to the preset number of edge cleaning cycles is not limited in this embodiment.

[0450] For example, when it is determined that the window cleaning robot will perform an edge-to-edge supplementary cleaning strategy, the starting boundary for edge-to-edge supplementary cleaning can be determined based on the acquired boundary type information. For example, a framed boundary can be selected as the starting boundary, or the left boundary can be selected as the starting boundary according to a preset priority. Further, the window cleaning robot is controlled to move sequentially along the starting boundary in a first preset direction to supplement the cleaning of the edge areas of the surface to be cleaned. During the movement, the window cleaning robot can maintain close contact with the boundary to ensure that the edge areas are thoroughly cleaned.

[0451] Furthermore, the number of times the window cleaning robot performs supplementary cleaning along the edges (i.e., the number of complete edge cleaning loops) is recorded. After each complete edge cleaning loop, it is determined whether the current count has reached the preset edge cleaning count. When it is determined that the number of supplementary edge cleaning loops has reached the preset count, the window cleaning robot is controlled to execute the return-to-starting-point strategy, returning to the starting position and ending the cleaning task.

[0452] Starting from the initial boundary, to form a complete closed loop, supplementary cleaning is performed sequentially in the order of left boundary, top boundary, right boundary, and bottom boundary. If the remaining boundary is insufficient to support a complete loop of supplementary cleaning, for example, only the bottom edge can be cleaned, then the process will switch to only perform supplementary cleaning on the bottom edge, or the current round of supplementary cleaning will end directly.

[0453] In this way, by controlling the window cleaning robot to move along the edges and recording the number of times it moves, the system ensures that the edge areas are cleaned at least a preset number of times, avoiding residual dirt due to insufficient cleaning and guaranteeing thorough cleaning of the edge areas. Furthermore, by clearly defining the starting boundary and the first preset direction, the edge cleaning path becomes more standardized and predictable, reducing path confusion or repeated coverage and improving the standardization of edge cleaning paths. Moreover, by setting an automatic return-to-starting-point strategy after completing supplementary edge cleaning, the system ensures the window cleaning robot safely returns to its starting position, avoiding inconvenience for users retrieving the robot due to uncertain task completion locations.

[0454] Furthermore, this application achieves automated control of edge cleaning by recording the number of times the window cleaning robot performs supplementary edge cleaning and judging the preset number of edge cleaning cycles, without requiring user intervention and improving the user experience.

[0455] Optionally, the method also includes:

[0456] Obtain the number of edge loops and / or edge entry conditions corresponding to the execution of the edge supplementary cleaning strategy;

[0457] The window cleaning robot is controlled to perform a supplementary cleaning strategy along the edges based on the number of edge loops and / or edge entry conditions.

[0458] In this embodiment, the number of edge-to-edge rotations refers to the total number of complete rotations the window cleaning robot makes while moving and cleaning along the perimeter of the surface to be cleaned during the execution of the edge-to-edge supplementary cleaning strategy. The number of edge-to-edge rotations determines the intensity of the edge-to-edge cleaning.

[0459] It should be noted that the number of circles along the edge can be determined in advance, by user settings, or automatically based on the cleaning mode; this application embodiment does not impose specific limitations on this.

[0460] Edge entry conditions refer to a set of preset conditions that must be met to trigger a window cleaning robot to perform edge supplementary cleaning. Edge entry conditions are used to determine whether edge supplementary cleaning needs to be performed.

[0461] Optionally, the edge entry condition is determined based on the following method:

[0462] Entry conditions are determined based on the cleaning mode: the entry conditions are met under deep cleaning mode, but not under quick cleaning mode.

[0463] Entry conditions are determined based on boundary type: the entry conditions are met in scenarios with framed boundaries, but not in scenarios without framed boundaries.

[0464] Based on user instructions: The entry conditions are met when the user manually initiates supplementary cleaning along the edge via the APP.

[0465] Based on the current task type: the entry condition is met after the global cleaning task is completed, but not after the local cleaning task is completed.

[0466] It should be noted that the edge entry condition can be a single condition or a combination of multiple conditions (such as deep cleaning mode and framed boundaries). The edge supplementary cleaning strategy is triggered only when all conditions are met. The specific content corresponding to the edge entry condition is not limited in the embodiments of this application.

[0467] Optionally, the entry conditions along the edge shall include at least the following six categories:

[0468] The main cleaning process is complete, meaning the main coverage path execution is finished, and the task enters the post-processing stage.

[0469] The current task allows for edge-supplementary cleaning, such as global cleaning tasks, zone cleaning tasks, or independent edge-supplementary cleaning tasks.

[0470] The current mode is configured with a greater than 0 edge cleaning cycle count. The deep cleaning mode and the quick cleaning mode each use their own edge cleaning cycle count configuration. Edge cleaning supplement is only allowed when the configured cycle count is greater than 0.

[0471] The main path type allows for edge-based supplementary cleaning. For example, in partition cleaning tasks and global cleaning tasks, edge-based supplementary cleaning is allowed if the main path is a Z-shaped path; while edge-based supplementary cleaning is generally not allowed for N-shaped paths.

[0472] The current boundary state satisfies the edge condition, that is, there exists at least the boundary required to execute the target edge, and it does not belong to the scenario of obviously borderless or insufficient boundary constraints.

[0473] The process either fails to enter the no-boundary protection or directly returns to the origin branch. If no effective border is detected, the width of the surface to be cleaned is too small, or a critical boundary is missing, edge-based supplementary cleaning is not allowed; instead, a protection strategy or a return-to-origin strategy is executed.

[0474] Optionally, different numbers of edge loops or different edge entry conditions can be set for deep cleaning mode and quick cleaning mode respectively, in order to balance coverage effect and work efficiency.

[0475] For example, before or during the decision-making process of the window cleaning robot to execute the edge-cleaning supplementary strategy, a preset number of edge-cleaning circles and / or edge-entry conditions can be obtained. Then, based on the edge-entry conditions, it can be determined whether the triggering requirements for executing edge-cleaning supplementary cleaning are met. For example, the entry conditions are met in deep cleaning mode, but not in quick cleaning mode.

[0476] Furthermore, after determining that the edge entry conditions are met, the window cleaning robot is controlled to execute an edge supplementary cleaning strategy based on a comprehensive judgment of the edge entry and edge circle count.

[0477] In this way, by setting the number of cleaning cycles along the edge, users or window cleaning robots can flexibly adjust the intensity of edge cleaning according to their cleaning needs, achieving precise control over edge cleaning and thus balancing cleaning effectiveness and time consumption. Furthermore, by setting the edge entry conditions, different triggering conditions can be met, making the triggering of supplementary edge cleaning more intelligent and flexible. Moreover, by judging the edge entry conditions, this step can be skipped in scenarios where edge cleaning is not needed, avoiding unnecessary edge cleaning.

[0478] In addition, users can set the number of rings along the edge and entry conditions to achieve personalized cleaning configurations and improve the user experience.

[0479] Optionally, obtain the surface features of the surface to be cleaned by the window cleaning robot, including:

[0480] The window cleaning robot is controlled to move along a second preset direction to the third target boundary of the surface to be cleaned, and performs edge probing and edge-adhering actions to obtain the surface features of the surface to be cleaned by the window cleaning robot.

[0481] In this embodiment, the second preset direction refers to the preset direction in which the window cleaning robot moves along the surface to be cleaned when performing edge probing and edge-adhering actions. The second preset direction can be horizontal (left-right direction) or vertical (up-down direction).

[0482] It should be noted that the second preset direction is selected based on the boundary of the target to be detected. For example, if the left or right boundary needs to be detected, then lateral movement is selected. If the upper or lower boundary needs to be detected, then longitudinal movement is selected.

[0483] It should also be noted that the first preset direction is usually used for the direction of movement (clockwise or counterclockwise) when cleaning along the edge, while the second preset direction is used for the direction of movement (horizontal or vertical) when probing and touching the edge. These are different concepts.

[0484] The third target boundary refers to the specific boundary of the surface to be cleaned that the window cleaning robot needs to reach and probe when performing edge probing and contacting actions. The third target boundary can be any boundary of the surface to be cleaned, including the upper boundary, lower boundary, left boundary, or right boundary.

[0485] It should be noted that the specific boundary chosen for the third objective depends on the requirements of the path planning. For example, before starting a lateral strip path, it is usually necessary to detect the left and right boundaries first. Before starting a longitudinal strip path, it is usually necessary to detect the upper and lower boundaries first.

[0486] It should also be noted that the initial boundary is the starting position during edge cleaning, while the third target boundary is the target position for edge probing and edge-hugging actions; these are different concepts.

[0487] Edge probing refers to the sequence of actions performed by a window cleaning robot as it approaches the boundary of the surface to be cleaned, using sensors and motion control to precisely detect the location and type of the boundary. Edge probing is a crucial step in acquiring surface features.

[0488] Optionally, the edge exploration action may include:

[0489] Reduce movement speed when approaching the boundary to avoid collision impact;

[0490] Wait for the boundary detection sensor (such as impact plate, ball head, infrared, ultrasonic, etc.) to trigger and confirm the boundary position;

[0491] Record the body's position coordinates when the sensor is triggered, as the boundary position;

[0492] Determine the boundary type (framed or frameless) based on the type of signal triggered by the sensor (e.g., impact trigger, ball head detachment).

[0493] It should be noted that the edge probing action is used to accurately obtain the location and type information of the boundary, providing data support for subsequent edge-fitting actions and path planning.

[0494] The edge-fitting action refers to the sequence of movements by which the window cleaning robot, after completing the edge-probing action, fine-tunes its position and posture to ensure close contact with the boundary. This edge-fitting action is a crucial step in ensuring accurate starting position for path planning.

[0495] Optionally, the edge-fitting action may include:

[0496] Move along the boundary direction to align the fuselage edge with the boundary;

[0497] Rotate the fuselage so that it is parallel or perpendicular to the boundary (depending on path planning requirements).

[0498] Sensors confirmed that the body and the boundary were tightly fitted without any gaps.

[0499] It should be noted that the edge-fitting action is used to ensure that the window cleaning robot starts the subsequent cleaning path with an accurate starting position and posture, avoiding path deviation or omissions due to deviation in the starting position.

[0500] For example, before the cleaning task begins, the control unit of the window cleaning robot first controls the robot to move along a second preset direction until it reaches the third target boundary of the surface to be cleaned. Upon reaching the vicinity of the third target boundary, it performs an edge-probing action to accurately detect the boundary's position and type. After completing the edge-probing action, it performs an edge-fitting action to ensure the robot's body fits tightly against the boundary, guaranteeing the accurate starting position of the subsequent cleaning path.

[0501] Furthermore, by performing edge probing and edge-fitting actions, and by using an edge detection unit, a collision detection unit, and a position estimation unit, the surface features of the surface to be cleaned are obtained. For example, the surface features may include: whether there is a solid border at the top of the current window, whether there are solid borders on the left and right sides of the current window, the actual horizontal width of the current window, whether the current window belongs to a narrow window scene, and whether the current window belongs to a frameless boundary scene.

[0502] In this way, by controlling the window cleaning robot to perform edge probing actions, the boundary position and type can be accurately detected, avoiding feature information deviations caused by sensor errors or irregularities in the surface to be cleaned, thus improving the accuracy of surface feature acquisition. By controlling the window cleaning robot to perform edge-fitting actions, the robot body is made to fit tightly against the boundary, ensuring the accuracy of the starting position of the subsequent cleaning path, avoiding path deviations or omissions caused by starting position deviations, and ensuring the accuracy of the starting position of path planning. Therefore, using edge probing and edge-fitting actions as important preliminary steps in path planning, by obtaining accurate boundary information, provides accurate input parameters for subsequent path strategy selection.

[0503] Furthermore, for windows with complex boundary types (such as partially framed and partially frameless) or irregular shapes, more accurate local feature information can be obtained by actively performing edge probing and edge-fitting actions, thereby enhancing the adaptability to complex surfaces to be cleaned.

[0504] Optionally, the method also includes:

[0505] During the cleaning operation of the window cleaning robot on the surface to be cleaned, if an interruption is detected, the breakpoint information is recorded. The breakpoint information includes at least the surface width information corresponding to the surface to be cleaned that the window cleaning robot was adsorbed at the time of the interruption.

[0506] When it is determined that the window cleaning robot has resumed cleaning operations, the current surface characteristics of the surface to be cleaned by the window cleaning robot are reacquired.

[0507] Based on the current surface features and breakpoint information, determine whether to execute the breakpoint continuation cleaning strategy.

[0508] It should be noted that when the window cleaning robot restarts and prepares to execute the breakpoint resume cleaning strategy, the window cleaning robot does not directly follow the historical breakpoint recovery path. Instead, it first re-executes edge probing, edge contact, and width measurement, and compares the current surface features of the surface to be cleaned with the surface features recorded in the historical breakpoint.

[0509] When the comparison result meets the preset similarity criterion, that is, it is determined that the current surface to be cleaned and the surface to be cleaned corresponding to the breakpoint are the same surface to be cleaned, then the breakpoint continuation cleaning process is allowed; otherwise, it is determined that the current surface to be cleaned is not the surface to be cleaned at the original breakpoint, then the breakpoint recovery state is cleared, and the planning is restarted according to the new global cleaning task or the partition cleaning task.

[0510] In this embodiment, breakpoint information refers to a set of data recorded by the window cleaning robot when the cleaning process is interrupted, used to identify the interruption location and status. Breakpoint information is the basic data for executing the breakpoint continuation cleaning strategy.

[0511] Optionally, breakpoint information includes at least one of the following:

[0512] Surface width information, i.e., the surface width of the area where the window cleaning robot is located when the interruption occurs, is used for subsequent width matching judgment;

[0513] Position coordinates, that is, the precise position of the window cleaning robot on the window surface when an interruption occurs (such as coordinates relative to the starting point).

[0514] The percentage of cleaned areas refers to the area or path that has been cleaned when the interruption occurs.

[0515] Current task type, i.e., the type of task being executed at the time of the interruption;

[0516] Timestamp, which is the time when the interruption occurred;

[0517] It also includes the number of path switches performed, the distance of the current breakpoint from the left boundary, the distance of the current breakpoint from the upper boundary, whether breakpoint continuation is allowed, and the current breakpoint recovery flag.

[0518] It should be noted that breakpoint information is usually stored in the non-volatile memory of the window cleaning robot to ensure that it is not lost after a power outage. This application does not specifically limit the storage method of breakpoint information. For example, breakpoint information can also be stored in the cloud.

[0519] The current surface features refer to the feature information of the currently adsorbed surface to be cleaned, which the window cleaning robot reacquires when resuming the cleaning operation. These current surface features are used for matching and judgment with breakpoint information.

[0520] Optionally, the current surface feature may include: current surface width information, current boundary type information, current boundary position, and current boundary combination state. This application embodiment does not limit the specific content corresponding to the current surface feature.

[0521] Optionally, the current surface features can be obtained by re-executing the edge probing and edge-adhering actions, or by real-time measurement using sensors. For example, after determining that the window cleaning robot has resumed cleaning operations, the robot is controlled to move along a second preset direction to the third target boundary of the surface to be cleaned, and the edge probing and edge-adhering actions are performed to obtain the current surface features of the surface to be cleaned that the robot has adhered to.

[0522] The breakpoint resume cleaning strategy refers to the path planning strategy of a window cleaning robot that, after an interruption is resumed, continues to execute the unfinished cleaning task from the breakpoint position based on the matching result between the breakpoint information and the current surface features.

[0523] It should be noted that the core objective of the breakpoint resume cleaning strategy is to continue the cleaning task with minimal repetitive coverage and in the shortest possible time after the interruption is restored.

[0524] Optionally, the execution process of the breakpoint-resume cleaning strategy includes:

[0525] Based on the position coordinates in the breakpoint information, move the window cleaning robot to the vicinity of the breakpoint location.

[0526] Starting from the breakpoint, continue executing the path planning strategy that was in effect before the interruption.

[0527] During the cleaning process, the boundary between cleaned and uncleaned areas is checked in real time to ensure that nothing is missed.

[0528] It should be noted that the breakpoint continuation cleaning strategy is only executed if the current surface features match the breakpoint information (e.g., the width is the same). If they do not match, the breakpoint continuation cleaning strategy is abandoned, and a new cleaning task is started, i.e., S201-S202 is executed.

[0529] For example, during the cleaning operation of the window cleaning robot, if an interruption is detected (such as battery depletion, user manual pause, fall risk triggering, communication interruption, etc.), the window cleaning robot can immediately record the breakpoint information. Furthermore, when it is determined that the window cleaning robot resumes the cleaning operation (such as after power is restored, the user clicks the "continue" button, or it is re-adhered to the window surface), the window cleaning robot re-acquires the current surface characteristics of the currently adhering surface to be cleaned.

[0530] Furthermore, based on the current surface features and breakpoint information, it is determined whether to execute the breakpoint continuation cleaning strategy. The determination logic may include at least one of the following:

[0531] Compare the current surface width with the width recorded at the breakpoint to see if they match or are within a certain error range. If they match, the window cleaning robot is considered to be still on the same window surface or in the same area, and the breakpoint cleaning can be resumed.

[0532] By combining other features, such as boundary type and the proportion of cleaned area, we can further confirm whether the robot is near the breakpoint.

[0533] If the judgment result is yes, the breakpoint resume cleaning strategy is executed, and the window cleaning robot is controlled to continue cleaning from the breakpoint position; if the judgment result is no, the breakpoint resume cleaning strategy is abandoned, a new cleaning task is restarted, or the user is prompted to confirm manually.

[0534] In this way, by recording the breakpoint information when the window cleaning robot is interrupted and intelligently matching it upon resumption, the robot can continue cleaning from the breakpoint position after the interruption is resolved, instead of starting from the beginning. This avoids repeated coverage of already cleaned areas and significantly improves cleaning efficiency. Furthermore, users do not need to manually replan or wait for the robot to re-clean the entire surface after an interruption, reducing user waiting time and operational steps, and enhancing the product's usability and intelligence.

[0535] Therefore, when a user removes the window cleaning robot from one surface to be cleaned and places it on another, the robot can automatically identify whether the surface has changed by using width matching. This allows it to decide whether to execute the breakpoint resume cleaning strategy, avoiding erroneous resume cleaning on different surfaces. Furthermore, by recording breakpoint information, even in the event of an unexpected interruption (such as a power outage or a fall), the robot can quickly locate and continue its task after recovery, enhancing its ability to handle abnormal situations.

[0536] Optionally, the breakpoint information also includes the boundary type information and boundary combination state of the surface to be cleaned by the window cleaning robot when the interruption occurs; based on the current surface features and breakpoint information, it is determined whether to execute the breakpoint continuation cleaning strategy, including:

[0537] If the current surface features and breakpoint information meet preset judgment conditions, the window cleaning robot is controlled to execute a breakpoint continuation cleaning strategy; the preset judgment conditions include at least one of the following:

[0538] The difference between the current surface width information and the surface width information in the breakpoint information is less than the preset difference threshold;

[0539] The current surface width is greater than or equal to the preset minimum width threshold;

[0540] The boundary type information and boundary combination status are consistent with the current boundary type information and current boundary combination status in the current surface features.

[0541] In this embodiment of the application, the preset judgment condition may refer to a preset set of conditions used to determine whether the current surface features match the breakpoint information, thereby deciding whether to execute the breakpoint continuation cleaning strategy.

[0542] Optionally, the preset judgment condition can be a combination of one or more conditions. This application does not specifically limit this; for example, the preset judgment condition may include:

[0543] Width difference matching: The difference between the current width and the breakpoint width is less than the preset difference threshold;

[0544] Width absolute value matching: The current width is greater than or equal to the preset width threshold;

[0545] Boundary type and combination state matching: The current boundary type and combination state are consistent with the breakpoint information.

[0546] The preset difference threshold can refer to the preset allowable error range used to determine whether the current surface width information is consistent with the surface width information in the breakpoint information.

[0547] It should be noted that the preset difference threshold is used to handle width measurement differences caused by sensor errors, minor window surface deformation, or positional offset of the window cleaning robot. For example, if the preset difference threshold is 5mm, then if the difference between the current width and the breakpoint width is less than 5mm, it is considered a width match. If the difference is greater than or equal to 5mm, it is considered a width mismatch.

[0548] Therefore, the preset difference threshold can be determined based on the sensor's measurement accuracy and the flatness of the surface to be cleaned. In this embodiment, the specific value corresponding to the preset difference threshold is not limited.

[0549] In this step, the preset minimum width threshold refers to a pre-defined width threshold used to determine whether the current surface width is sufficient to support the safe execution of the breakpoint continuation cleaning strategy. When the current surface width of the surface to be cleaned is less than the preset minimum width threshold, the window cleaning robot cannot safely execute the continuation cleaning path on that surface, and therefore abandons continuation cleaning even if the width matches.

[0550] For example, if the preset minimum width threshold is 200mm, then if the current width is 150mm, the condition is not met, and the breakpoint resume cleaning strategy will not be executed. If the current width is 250mm, the condition is met, and the breakpoint resume cleaning strategy can be executed.

[0551] It should be noted that window-to-window determination is only allowed when the current surface width is greater than a preset minimum width threshold. Therefore, the preset minimum width threshold can be related to the width of the window cleaning robot's body, the minimum safe turning radius, or the effective coverage width of the cleaning module. The preset width threshold may be the same as or different from the preset width threshold in the above embodiments, and this application does not specifically limit it.

[0552] For example, when a window cleaning robot's cleaning task is interrupted, in addition to recording the surface width information, it can also record the boundary type information and boundary combination state of the surface to be cleaned at the time of the interruption. When the window cleaning robot resumes cleaning, it can reacquire the current surface features of the currently adsorbed surface to be cleaned, including the current surface width information, the current boundary type information, and the current boundary combination state. Then, based on the current surface features and the breakpoint information, it determines whether a preset judgment condition is met. If the preset judgment condition is met (it can be a single condition or multiple conditions met simultaneously), the window cleaning robot is controlled to execute the breakpoint continuation cleaning strategy; otherwise, the breakpoint continuation cleaning strategy is abandoned, and a new cleaning task is started.

[0553] Thus, by introducing breakpoint information and more refined judgment conditions—namely, by using current surface width information, boundary type information, and boundary combination state—this application can more accurately determine whether the window cleaning robot is still on the same surface or in the same area to be cleaned, avoiding misjudgments caused by coincidental width matching and improving the accuracy of resuming cleaning after a breakpoint. Furthermore, the preset judgment conditions support the execution of the resuming cleaning strategy as long as at least one condition is met, allowing the window cleaning robot to select the appropriate judgment logic according to the actual scenario, thereby providing flexible judgment logic.

[0554] It should also be noted that when a user removes a window cleaning robot from one surface to be cleaned and places it on another, even if the widths happen to be the same, the boundary type information and boundary combination state may differ. By matching the aforementioned preset judgment conditions, it is possible to accurately identify whether the surface to be cleaned has been changed, avoiding the incorrect execution of continued cleaning on different surfaces.

[0555] Furthermore, by setting the current surface width information to be greater than or equal to the preset minimum width threshold, the risk of performing follow-up cleaning on extremely narrow surfaces to be cleaned can be eliminated, because extremely narrow surfaces to be cleaned may not be able to safely perform follow-up cleaning paths.

[0556] Optionally, the method also includes:

[0557] If the current surface features and breakpoint information do not meet the preset judgment conditions, the corresponding path planning strategy is selected based on the current surface features, and the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

[0558] For example, in the decision-making process of a window cleaning robot to determine whether to execute a breakpoint-resume cleaning strategy based on current surface features and breakpoint information, if the judgment result does not meet the preset judgment conditions, the execution of the breakpoint-resume cleaning strategy is abandoned. Instead, based on the current surface features, a corresponding path planning strategy is reselected according to the conventional path planning strategy selection logic. For example, if the current surface is a wide window with frames on both sides, a full-coverage path strategy is selected. If the current surface is a narrow window or frameless, a protection strategy or a partial strip coverage strategy is selected. If the current surface is a local area, a first strategy or a partial strip coverage strategy is selected based on the local features. For specific details, please refer to the description of the above embodiments, which will not be repeated here.

[0559] Furthermore, based on the newly selected path planning strategy, the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned, which is equivalent to starting a completely new cleaning task.

[0560] Thus, if a pre-defined judgment condition determines that the surface to be cleaned has been replaced or its characteristics have changed, forcibly executing a breakpoint-based cleaning strategy may lead to path errors, duplicate coverage, or omissions. By abandoning the breakpoint-based cleaning strategy and reselecting a path planning strategy, efficiency losses caused by incorrect cleaning continuation are avoided. Moreover, in cases where cleaning cannot be continued, reselecting a path planning strategy and executing the cleaning operation ensures that the current surface to be cleaned is still completely cleaned, preventing uncleaned areas from being left due to interruption and ensuring the integrity of the cleaning task.

[0561] Furthermore, because the window cleaning robot can automatically adjust its path planning strategy based on the current surface characteristics, no user intervention is required. Even if the surface conditions change after an interruption, the window cleaning robot can quickly adapt and start a new cleaning task, enhancing its adaptability. Moreover, the above process eliminates the need for users to manually determine whether to execute a resume cleaning strategy or restart the path planning strategy; the window cleaning robot automatically completes the decision-making and switching, simplifying user operation and improving the product's intelligence level.

[0562] Optionally, the breakpoint information also includes the window cleaning robot's position information at the time of the interruption, controlling the window cleaning robot to execute the breakpoint continuation cleaning strategy, including:

[0563] Starting with the location information, the window cleaning robot continues to execute the corresponding path planning strategy to perform cleaning operations on the surface to be cleaned.

[0564] In this embodiment, location information refers to the precise coordinates or relative position description of the window cleaning robot on the surface to be cleaned when an interruption occurs. Location information is used to identify the spatial location at the time of the interruption.

[0565] Optionally, location information may include:

[0566] Absolute coordinates: X and Y coordinates relative to the starting point or reference point of the surface to be cleaned;

[0567] Relative coordinates: the offset relative to the boundary of the surface to be cleaned (such as the top boundary or left boundary);

[0568] Path progress: The completion percentage of the current path or the number of the path segment that has been executed.

[0569] It should be noted that the specific content corresponding to the location information in this application embodiment is not limited. The location information can be calculated and recorded in real time through the window cleaning robot's odometer, inertial measurement unit, boundary sensor or visual positioning system.

[0570] Optionally, the location information can be stored together with the breakpoint information in non-volatile memory to ensure that it is not lost after power failure. This application embodiment does not specifically limit the storage method of the location information. For example, it can also be stored in the cloud.

[0571] For example, if it is determined that the window cleaning robot is executing a resume cleaning strategy, the position information of the window cleaning robot at the time of the interruption can be extracted from the recorded breakpoint information. Furthermore, the window cleaning robot is controlled to move to the coordinate position indicated by this position information, serving as the starting point for resume cleaning. Then, the window cleaning robot continues to execute the corresponding path planning strategy (such as a horizontal strip path, a vertical strip path, an edge cleaning path, etc.) that was being executed before the interruption, to continue the cleaning operation on the surface to be cleaned.

[0572] It should be noted that when the window cleaning robot restores its path, it can automatically handle the path connection problem, ensuring that the path starting from the break point is smoothly connected to the path before the interruption, avoiding repeated coverage or omissions.

[0573] In this way, by utilizing location information to precisely resume cleaning from the breakpoint, the window cleaning robot can accurately return to the position where the interruption occurred, rather than restarting from the beginning of the entire surface to be cleaned. This minimizes the area of ​​repeated cleaning and achieves precise resume cleaning from the breakpoint. Since cleaning only needs to continue from the breakpoint, instead of re-cleaning the entire surface, cleaning time can be significantly saved, especially when the surface to be cleaned is large or most of the area has already been cleaned, thus improving cleaning efficiency.

[0574] Furthermore, by using location information for path connection processing, the continuity and integrity of the restored path are ensured to be consistent with the path before the interruption, avoiding coverage omissions or duplications caused by path misalignment.

[0575] Furthermore, the above process does not require users to manually record the interruption location or replan the path. The window cleaning robot can automatically complete the location positioning and path restoration, which greatly improves the ease of use and intelligent experience of the product.

[0576] Optionally, the method also includes:

[0577] During the cleaning operation of the window cleaning robot on the surface to be cleaned, the feedback information of various sensors in the window cleaning robot is detected;

[0578] When feedback indicates that the window cleaning robot has encountered an anomaly or boundary anomaly, the robot is controlled to correct its path planning strategy so that it can perform cleaning operations based on the corrected path planning strategy.

[0579] In this embodiment, the feedback information refers to data collected and reported in real time by various sensors in the window cleaning robot during the cleaning process, reflecting the robot's own status and the external environment. Optionally, the feedback information may include:

[0580] Information detected by the boundary detection sensor, such as the impact trigger signal, the ball head separation signal, and the infrared distance value;

[0581] Information detected by the drop sensor, such as the drop sensor trigger signal;

[0582] Information detected by the motor current sensor, such as the motor current value (used to detect stall).

[0583] Information detected by attitude sensors, such as gyroscope / accelerometer data (used to detect tilt and collisions);

[0584] Information detected by the suction sensor, such as the suction value (used to detect a decrease in adsorption force).

[0585] It should be noted that the specific content of the feedback information in this application embodiment is not limited; the feedback information is used to determine whether an anomaly has occurred. For example, the feedback information may also include: a boundary presence marker, a virtual boundary marker, the current cleaned width, the number of path switching times, the current operation stage, etc.

[0586] It should also be noted that the abnormal situation of the window cleaning robot or encountering boundary abnormality refers to the abnormal situation that occurs during the cleaning process due to changes in its own state or the external environment, which requires the correction of the path planning strategy.

[0587] Optionally, common abnormal situations include: robot-specific abnormalities, boundary abnormalities, and other abnormalities. This application does not limit the specific scenarios corresponding to these abnormal situations.

[0588] Among them, robot malfunctions can include motor stall (abnormally high motor current, indicating that the cleaning wheel or drive wheel is stuck), decreased suction (the suction sensor detects that the suction force is lower than the safety threshold, indicating a risk of falling), abnormal posture (the gyroscope detects that the body tilt angle exceeds the preset value, indicating that it may have encountered an obstacle or the surface to be cleaned is uneven), body collision, and sensor failure (a key sensor has no feedback or abnormal feedback).

[0589] Boundary anomalies can include boundary type mutations (suddenly changing from a framed boundary to an unframed boundary), boundary missing (no boundary or risk of an unframed bottom edge is detected at a location where a boundary was expected), and boundary location mutations.

[0590] Other anomalies may include communication interruption (the communication connection between the window cleaning robot and the APP is interrupted), insufficient battery power (the battery power is lower than the safety threshold, and the task needs to be ended in advance), user intervention (the user manually presses the pause or stop button), etc.

[0591] For example, during the cleaning operation of the window cleaning robot, feedback information from various sensors within the robot is continuously monitored. Based on this feedback, it is determined whether the robot has encountered any abnormalities or boundary anomalies. These include motor stalling, decreased suction, abnormal posture, sensor failure, or sudden changes in boundary type (e.g., from framed to frameless), missing boundaries, or abrupt changes in boundary position. Furthermore, upon identifying an anomaly or boundary anomaly, the robot is controlled to correct its current path planning strategy. Finally, the robot continues cleaning operations based on the corrected path planning strategy until the task is completed or another anomaly is encountered.

[0592] Optionally, the correction method may include:

[0593] Change the direction of movement, step distance, or path type (e.g., switch from a horizontal strip to a vertical strip);

[0594] Perform safety maneuvers such as reversing, slowing down, stopping, or re-testing the edge;

[0595] Switch from the current path planning strategy to a protection strategy, a local strip coverage strategy, or a border cleaning strategy.

[0596] It should be noted that the embodiments of this application do not specifically limit the modification method; the above are merely illustrative examples.

[0597] Optionally, in the event of an anomaly or boundary anomaly encountered by the window cleaning robot, the robot can be controlled to perform at least one of the following actions: pause, retreat, re-switch path, recalculate boundary, switch to edge-to-edge supplementary cleaning, execute return-to-start strategy, or end the current task.

[0598] In this way, this application improves the safety of the cleaning process by monitoring the window cleaning robot in real time for any abnormalities, promptly correcting its path, or executing safety actions to prevent damage or accidents. Furthermore, by adjusting the path planning strategy in real time, the window cleaning robot can cope with various emergencies, reducing cleaning interruptions or task failures caused by abnormalities and improving task completion rates. This dynamic correction mechanism makes the path planning strategy more robust, adapting to changes in the conditions of different types of surfaces to be cleaned, ensuring complete cleaning coverage.

[0599] Furthermore, when encountering boundary anomalies (such as irregular window shapes or abrupt changes in boundary types), the window cleaning robot can continue to complete the cleaning task by dynamically correcting the path planning strategy, instead of stopping or failing directly, thus enhancing the adaptability of the window cleaning robot to complex surfaces to be cleaned.

[0600] Optionally, the window cleaning robot performs cleaning operations based on a modified path planning strategy, including any of the following:

[0601] Control the window cleaning robot to pause its cleaning operation;

[0602] After controlling the window cleaning robot to retreat a preset distance, it performs the cleaning operation again based on the path planning strategy;

[0603] After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy.

[0604] Control the window cleaning robot to execute the edge cleaning strategy.

[0605] In this embodiment, the preset distance may refer to the fixed distance value of the window cleaning robot when it performs the back-and-retry correction method.

[0606] It should be noted that the preset distance is a fixed value used to control the extent of the backward movement. Different preset distances can be set for different scenarios. For example, for short backward movement scenarios to avoid minor obstacles or boundary anomalies, the preset distance can be set to 5cm. For medium backward movement scenarios to avoid moderate obstacles or anomalies, the preset distance can be set to 10cm. For long backward movement scenarios to avoid large areas of anomalies, the preset distance can be set to 20cm.

[0607] Optionally, the preset distance can be determined based on factors such as the robot's body size, cleaning module width, and sensor detection range to ensure that it can avoid abnormal areas after retreating, while not retreating too far and causing a decrease in cleaning efficiency. This application embodiment does not limit the specific value corresponding to the preset distance; the above is merely an example.

[0608] For example, after the window cleaning robot detects an anomaly or boundary anomaly, it can select one or a combination of the following four correction methods based on the anomaly type and severity:

[0609] Method 1: Control the window cleaning robot to immediately pause the current cleaning operation and wait for user intervention or automatic resumption (such as waiting for suction to recover or waiting for the sensor to reset).

[0610] Method 2: Control the window cleaning robot to retreat a preset distance, and then perform the cleaning operation again based on the original path planning strategy or the adjusted path planning strategy. This method is often used when encountering minor obstacles or boundary abnormalities, and to avoid abnormal areas by retreating.

[0611] Method 3: Control the window cleaning robot to abandon the current path and plan a new one, such as switching from a horizontal strip to a vertical strip, or from a global path to a local path, and then perform the cleaning operation based on the new path. This method is often used when encountering obstacles that cannot be bypassed or when there are abrupt changes in boundary types.

[0612] Method 4: Control the window cleaning robot to switch to the edge cleaning strategy, moving along the current boundary and cleaning the edge area. This method is often used when encountering boundary anomalies (such as abrupt changes in boundary type), ensuring that the edge area is covered through edge cleaning.

[0613] Furthermore, depending on the selected correction method, the window cleaning robot controls itself to perform corresponding actions, such as pausing, reversing, changing paths, and cleaning along edges. After the correction actions are completed, it continues the cleaning operation until the task is completed or another abnormality is encountered.

[0614] Different abnormal situations require different responses. Therefore, this application provides four correction methods: pause, reverse, path switching, and edge cleaning. This allows the window cleaning robot to select the appropriate correction action based on the type of abnormality, thus improving its flexibility in responding to situations.

[0615] Among these improvements, the corrective methods of retrying or switching paths allow the window cleaning robot to bypass minor obstacles or abnormal areas and continue cleaning, rather than stopping or failing directly, thus improving the robustness of the cleaning task. The corrective method of switching to an edge cleaning strategy ensures that edge areas are still cleaned even in cases of boundary abnormalities, preventing edge omissions and ensuring thorough edge coverage. The corrective method of pausing execution allows the window cleaning robot to stop promptly when encountering serious abnormalities (such as insufficient suction or sensor failure), preventing damage to the robot or safety accidents and improving operational safety.

[0616] Optionally, the method also includes:

[0617] During the cleaning operation of the window cleaning robot on the surface to be cleaned, if the window cleaning robot encounters abnormal working conditions, the type of abnormality is determined.

[0618] Based on the type of exception, the window cleaning robot selects the corresponding processing strategy so that the window cleaning robot can perform the corresponding operation based on the processing strategy.

[0619] In this embodiment, abnormal operating conditions can refer to any unexpected situation encountered by the window cleaning robot during the cleaning process that deviates from the normal operating state. For example, abnormal operating conditions may include the detection of a window handle, local obstacles, sudden changes in posture, unexpected boundaries, or other abnormal operating conditions. This embodiment does not limit the specific operating conditions corresponding to abnormal operating conditions.

[0620] Optionally, abnormal operating conditions can be detected in real time through sensor feedback information, motor current monitoring, communication status monitoring, etc. This application embodiment does not specifically limit the detection method.

[0621] An anomaly type can refer to a label used to classify abnormal operating conditions, distinguishing between anomalies of different natures. Anomaly types can be classified according to their source, severity, and scope of impact. This application does not specifically limit the basis and categories for classification.

[0622] Optionally, the anomaly type classification can include: safety anomalies, mechanical anomalies, environmental anomalies, communication anomalies, and electrical anomalies, etc.

[0623] Among them, safety-related anomalies can refer to anomalies that may endanger the safety of the window cleaning robot or cause it to fall, such as decreased suction or triggering of the fall sensor.

[0624] Mechanical anomalies can refer to anomalies involving jamming or damage to mechanical parts of the robot, such as motor stalling or cleaning wheels getting stuck.

[0625] Environmental anomalies can refer to anomalies caused by changes in the external environment, such as abrupt changes in boundary type, uneven window surface, detection of window handle, local obstacles, etc.

[0626] Communication-related anomalies can refer to anomalies involving interruption or abnormality of the communication link, such as unexpected network disconnection or network connection timeout.

[0627] Battery-related anomalies can refer to anomalies caused by insufficient battery power.

[0628] In this application, a processing strategy can refer to a set of pre-defined operation instructions or action sequences for a specific type of anomaly, used to deal with abnormal operating conditions. Different anomaly types correspond to different processing strategies. For example, safety-related anomalies correspond to safety processing strategies, which may include immediately stopping all movement, executing a return-to-start strategy, or issuing an alarm.

[0629] Mechanical anomalies correspond to mechanical recovery strategies, which can include reversing the motor, reversing a preset distance, or retrying.

[0630] For environmental anomalies, the corresponding path correction strategies can include switching path types, performing edge cleaning, or re-exploring the edge.

[0631] Communication-related exceptions correspond to communication recovery strategies, which can include re-establishing the network connection, retrying after a timeout, or switching the communication channel.

[0632] Power-related anomalies correspond to corresponding power management strategies, such as reducing cleaning speed, ending the task early, or returning to the charging dock.

[0633] Optionally, the processing strategies can be stored in the window cleaning robot's memory as a strategy library, so that the window cleaning robot can call the corresponding processing strategy from the strategy library according to the exception type.

[0634] For example, during the cleaning operation, the window cleaning robot continuously monitors for abnormal conditions. When an abnormal condition is detected, it is further analyzed and the type of abnormality is determined. Based on the determined abnormality type, the window cleaning robot selects a corresponding processing strategy from a pre-set strategy library. Furthermore, the robot is controlled to perform corresponding operations based on the selected processing strategy, such as stopping, reversing, changing paths, or reconnecting.

[0635] Thus, by introducing anomaly type classification and differentiated handling strategies, this application not only achieves precise anomaly response but also improves the efficiency of anomaly handling. Different anomaly types require different handling methods. Therefore, by identifying the anomaly type, appropriate handling strategies can be selected specifically, improving the accuracy of the response. Furthermore, for different types of anomalies, the window cleaning robot can quickly match preset handling strategies without complex comprehensive judgments, improving the response speed of anomaly handling. Moreover, through differentiated handling strategies, most anomalies can be automatically handled without disturbing the user, improving the user experience.

[0636] Furthermore, the design of the aforementioned differentiated handling strategies enhances the window cleaning robot's adaptability to various abnormal situations, thereby improving its robustness and safety. For example, in the case of safety-related anomalies, the window cleaning robot can prioritize the execution of safety handling strategies to ensure robot safety; in the case of mechanical anomalies, the window cleaning robot can attempt to resume operation to avoid task interruption.

[0637] Optionally, the window cleaning robot performs corresponding operations based on the processing strategy, including any of the following:

[0638] Control the window cleaning robot to perform local obstacle avoidance actions, and after obstacle avoidance, continue to perform cleaning operations based on the selected path planning strategy;

[0639] After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy.

[0640] Control the window cleaning robot to prevent it from continuing the breakpoint cleaning strategy;

[0641] Control the window cleaning robot to pause its cleaning operation;

[0642] Control the window cleaning robot to execute the return-to-start strategy.

[0643] In this embodiment of the application, local obstacle avoidance action can refer to a set of short-distance action sequences executed by the window cleaning robot when it encounters a local obstacle or abnormal area, in order to avoid the abnormal area and restore the cleaning path.

[0644] Optionally, local obstacle avoidance actions may include:

[0645] Move a certain distance (e.g., 5cm) in the opposite direction of the current direction;

[0646] Rotate by a preset angle (e.g., 30 degrees, 45 degrees) to change the direction of movement;

[0647] Move along the edge of the obstacle, bypass the obstacle, and then resume the original path;

[0648] Reduce movement speed and proceed slowly through the abnormal area.

[0649] It should be noted that the purpose of local obstacle avoidance is to avoid abnormal areas with minimal path adjustments and to restore the original path planning strategy as quickly as possible, thereby reducing the impact on cleaning efficiency. This application does not limit the specific operations corresponding to local obstacle avoidance; the above are merely illustrative examples.

[0650] For example, after detecting an abnormal operating condition and determining the type of abnormality, the window cleaning robot can select a corresponding processing strategy from the strategy library. This processing strategy will further specify the execution of one or a combination of the following five operations:

[0651] Method 1: Control the window cleaning robot to perform local obstacle avoidance actions (such as backing up, turning, and detouring). After avoiding obstacles or abnormal areas, continue to perform cleaning operations based on the currently selected path planning strategy.

[0652] Method 2: Control the window cleaning robot to abandon the current path and replan a new path (such as switching from a horizontal strip to a vertical strip, or from a global path to a local path), and then perform the cleaning operation based on the new path.

[0653] Method 3: Control the window cleaning robot to prevent it from continuing the resume cleaning strategy. This operation is usually used to prevent the robot from incorrectly resuming cleaning after encountering serious abnormalities (such as window replacement or sensor failure).

[0654] Method 4: Control the window cleaning robot to immediately pause the current cleaning operation and wait for user intervention or for the window cleaning robot to resume automatically.

[0655] Method 5: Control the window cleaning robot to execute the return-to-start strategy, that is, stop the current cleaning task and return to the starting position according to the preset path.

[0656] Furthermore, based on the instructions of the processing strategy, the window cleaning robot is controlled to perform corresponding operations.

[0657] Different types and severity of anomalies require different actions. Therefore, this application provides five methods—local obstacle avoidance, path switching, disabling interrupted cleaning, pausing execution, and returning to the starting point—to enable the window cleaning robot to select the appropriate operation based on the anomaly situation, thereby improving its flexibility in handling anomalies. Specifically, by using local obstacle avoidance or path switching, the window cleaning robot can bypass minor anomalies and continue cleaning, avoiding task interruption and improving the continuity of the cleaning task.

[0658] By prohibiting breakpoint-based resume cleaning, the window cleaning robot can avoid erroneous resume cleaning after the surface to be cleaned is changed or the sensor fails, which could lead to path confusion or missed coverage, effectively preventing incorrect resume cleaning. The pause or return-to-start strategy allows the window cleaning robot to stop or return promptly when encountering serious abnormalities (such as insufficient suction or risk of falling), ensuring the robot's safety.

[0659] In this way, the window cleaning robot can handle minor anomalies automatically without the user noticing; for serious anomalies, the window cleaning robot can notify the user by pausing execution or returning to the starting point, thereby optimizing the user experience.

[0660] Optionally, the method also includes:

[0661] After completing the cleaning operation, the corresponding return-to-start strategy is selected based on the path planning strategy and / or surface characteristics, and the window cleaning robot is controlled to return to the starting position based on the return-to-start strategy.

[0662] For example, after the window cleaning robot completes all cleaning tasks on the surface to be cleaned, it can select a corresponding return-to-start strategy from a preset return-to-start strategy library based on the path planning strategy used during the cleaning process and / or surface characteristics. For instance, the return-to-start strategy can be selected based on the path planning strategy: if the path planning strategy is the first strategy, it can choose to return directly along the current row direction or along the boundary. If the path planning strategy is the second strategy, it can choose to return directly along the current column direction or along the boundary. If the path planning strategy is an edge-cleaning strategy, it can choose to return in reverse along the boundary or along the shortest path.

[0663] In addition, the return-to-origin strategy can be selected based on surface features. For example, if the surface is wide, a Z-shaped path return-to-origin strategy can be chosen. If the surface is narrow, a boundary-following return-to-origin strategy can be chosen to avoid collisions. If the boundary type is frameless, a frameless return-to-origin strategy can be chosen to avoid the risk of falling.

[0664] In this way, the window cleaning robot is controlled to return to the starting point position based on the selected return-to-start strategy.

[0665] Thus, this application, by intelligently selecting a return-to-start strategy based on path planning and / or surface characteristics, not only improves the efficiency of the return-to-start process but also ensures its safety. Specifically, by selecting an appropriate return-to-start strategy, the return path can be shortened, return time reduced, and overall cleaning efficiency improved. Selecting a return-to-start strategy based on surface characteristics avoids the risk of collisions or falls due to improper path selection on narrow or frameless surfaces, ensuring a safe return process.

[0666] Furthermore, by associating the return-to-start strategy with the path planning strategy, this application can ensure that the return path is consistent with the cleaning path, avoiding path conflicts or confusion, optimizing the integrity of path planning, and through intelligent selection of the return-to-start strategy, the window cleaning robot can automatically, efficiently and safely return to the starting point without manual intervention from the user, thus improving the product's intelligence level and user experience.

[0667] Optionally, a corresponding return-to-start strategy can be selected based on the path planning strategy and / or surface features, including:

[0668] When the path planning strategy is a strategy with the longitudinal direction as the main direction of travel, the window cleaning robot is controlled to return to the starting point along the longitudinal direction.

[0669] When the path planning strategy is a strategy with the lateral direction as the main travel direction, the general strategy for controlling the window cleaning robot is to return to the starting point.

[0670] When the surface features indicate that the boundary of the surface to be cleaned is incomplete, the window cleaning robot adopts a frameless return-to-start strategy.

[0671] When the path planning strategy is an edge cleaning strategy, the current boundary and starting edge information are determined based on the surface features. After adjusting the body posture of the window cleaning robot based on the current boundary and starting edge information, the window cleaning robot is controlled to adopt a general return-to-starting-point strategy.

[0672] In this embodiment of the application, the strategy of returning to the starting point along the longitudinal direction can be understood as a narrow window adaptation return-to-origin strategy. It can refer to a return-to-origin strategy applicable to cleaning paths (such as N-type paths and narrow window strip coverage paths) with the longitudinal direction as the main direction of travel. The core is to retreat along the longitudinal direction to avoid large-scale lateral switching.

[0673] Optionally, the execution flow corresponding to the strategy of returning to the starting point in the vertical direction includes:

[0674] First, adjust the orientation of the window cleaning robot so that it is vertically upward.

[0675] Control the window cleaning robot to move upwards to detect the reference boundary (such as the upper boundary), and record the position at that moment as the starting point for retraction.

[0676] Control the window cleaning robot to retreat longitudinally from the starting point according to the previously recorded upper boundary distance.

[0677] The process ends after retracing to near the starting column of the original strip, returning to the starting position.

[0678] Therefore, the above-mentioned return route relies heavily on the upper boundary distance and avoids large lateral switching movements as much as possible. It is more suitable for larger surfaces to be cleaned or for return routes after strip-style round-trip coverage.

[0679] A general return-to-origin strategy (GROUP) can refer to a return-to-origin strategy applicable to clean paths where the main direction of travel is lateral (such as Z-shaped paths or paths used in ordinary wide-window scenarios). A general return-to-origin strategy typically employs a three-stage return process: longitudinal backtracking, lateral correction, and compensation when necessary.

[0680] For example, the execution flow corresponding to the general return-to-start strategy includes:

[0681] Control the window cleaning robot to move forward and probe the reference boundary in order to establish the return reference starting point.

[0682] Control the window cleaning robot to retreat longitudinally according to the previously recorded upper boundary distance, returning to a longitudinal position close to the original starting position, such as being on the same strip.

[0683] The control system will steer the aircraft into a lateral attitude and perform a lateral crossing by staying close to the left boundary based on the previously recorded distance to the left boundary or the nearest boundary determination. If a near-edge failure, collision, or boundary anomaly occurs during the lateral crossing, a protection strategy will be implemented and the aircraft will return to the starting position.

[0684] It should be noted that in this step, the key point of the frameless return-to-start strategy is that it no longer relies on the closed-loop return logic of ordinary entity borders, but instead uses the method of probing the edge-backing-turning-probing the edge again to gradually return to the vicinity of the starting position.

[0685] Optionally, the execution flow corresponding to the frameless return-to-start strategy may also include:

[0686] The window cleaning robot is first controlled to make a turning maneuver to avoid the risky direction;

[0687] Then, the robot's body posture is readjusted to a vertically upward orientation.

[0688] Control the window cleaning robot to move upwards along the edge and record the reference starting point;

[0689] Control the window cleaning robot to retreat downwards according to the recorded upper boundary distance;

[0690] The window cleaning robot is then controlled to turn to a lateral position to test the waters by moving to the side.

[0691] The probing continues until a safe lateral distance is reached, and a posture recovery can be performed again if necessary.

[0692] Control the window cleaning robot to repeat the above process until the return trip is completed.

[0693] It should be noted that adopting a frameless return-to-start strategy can avoid returning according to the logic of ordinary entity borders, thereby improving the security and success rate under frameless edge conditions.

[0694] The current boundary refers to the window surface boundary along which the window cleaning robot is currently moving while executing its edge cleaning strategy. The current boundary can include the top boundary, bottom boundary, left boundary, and right boundary.

[0695] The current boundary is used to determine the path direction for edge cleaning, providing a basis for adjusting the posture of the window cleaning robot during its return journey.

[0696] The starting edge information refers to the initial boundary selected by the window cleaning robot when it begins to execute its edge cleaning strategy, along with its corresponding position and orientation. Starting edge information can include: starting edge type, starting edge position, boundary state, and starting edge orientation.

[0697] The starting edge type indicates whether the starting edge is the top, bottom, left, or right boundary. The starting edge position refers to the specific coordinates or relative position of the starting edge on the surface to be cleaned. The starting edge orientation refers to the orientation of the window cleaning robot when it begins cleaning along the edge (e.g., vertically upward, horizontally to the left, etc.).

[0698] For example, Figure 3A flowchart illustrating a differentiated return-to-starting point method for a window cleaning robot, as provided in this application embodiment, is shown below. Figure 3 As shown, the process includes:

[0699] Step a: After the window cleaning robot has completed its main body cleaning, the path pattern and window type information (i.e., surface features) corresponding to the executed path planning strategy can be read. Further, it is determined whether it is an N-type path strategy or a narrow window path strategy. If so, a narrow window adapted return-to-origin strategy is adopted, that is, returning to the starting point in the longitudinal direction. This narrow window adapted return-to-origin strategy mainly uses the recorded upper boundary distance and strip switching state to control the window cleaning robot to return. If not, proceed to step b.

[0700] Step b: Determine if it is a frameless boundary. If yes, select the frameless return-to-start strategy. If no, avoid the risk of frameless edge return. Further, determine if it is a Z-shaped path strategy or a normal wide-window path strategy. If yes, adopt the general return-to-start strategy. If no, select the return strategy that matches the current task.

[0701] Step c: After executing the above return-to-start strategy, determine whether the return to the starting point was successful. If yes, end the cleaning task; otherwise, perform anomaly detection and handling. The specific process can be referred to the handling methods for anomalies or abnormal working conditions in the above embodiments, and will not be repeated here. Alternatively, if it is determined that the return to the starting point was unsuccessful, a protection strategy can also be executed.

[0702] In some embodiments, for the edge-cleaning strategy, the window cleaning robot executes a specific return-to-origin strategy after edge cleaning based on the current boundary and starting edge information. The execution process corresponding to the specific return-to-origin strategy after edge cleaning includes:

[0703] After the window cleaning robot completes edge cleaning or independently performs edge cleaning tasks, it records the current starting edge and starting edge information.

[0704] The window cleaning robot is controlled to perform special posture adjustments based on the starting edge information along the edge, so that the body switches from the posture at the end of the edge cleaning to a posture suitable for returning to the origin.

[0705] Then control the window cleaning robot to execute the general return-to-start strategy.

[0706] If the window cleaning robot detects a frameless or insufficient boundary during the edge cleaning process, it will execute a frameless return-to-start strategy.

[0707] It should be noted that the embodiments of this application do not limit the return-to-start strategy adopted in different scenarios; the above are merely illustrative examples.

[0708] Different cleaning paths and surface conditions require different return-to-start paths. Therefore, this application achieves differentiated return-to-start path planning by intelligently selecting different return-to-start strategies based on path planning strategies and surface characteristics. This allows the window cleaning robot to choose the appropriate return method according to the actual scenario. For example, by controlling the window cleaning robot to return to the starting point in the longitudinal direction, unnecessary lateral switching actions can be avoided in narrow window scenarios, reducing the return path length and time, and improving return efficiency.

[0709] By controlling the window cleaning robot to adopt a universal return-to-start strategy, the window cleaning robot can ensure the integrity and accuracy of the return path in wide window scenarios through a three-stage return process of "vertical retreat, lateral correction, and compensation when necessary".

[0710] By controlling the window cleaning robot to adopt a frameless return-to-start strategy, the window cleaning robot can use a more conservative "edge probe-reverse-turn-re-edge probe" method, avoiding the risk of falling due to reliance on physical frames and ensuring the safety of the frameless window surface during the return trip.

[0711] By controlling the window cleaning robot to adopt a dedicated return-to-origin strategy after cleaning along the edge, and by adjusting the posture based on the starting and ending postures of the edge cleaning, a smooth transition from edge cleaning to returning to the origin is ensured, avoiding path confusion caused by posture conflicts of the window cleaning robot.

[0712] Therefore, this application can select differentiated return-to-origin strategies based on different surface characteristics of the surface to be cleaned and / or different path planning strategies. For example, different return methods can be used for narrow windows, frameless windows, and different main path patterns. Compared with the unified return-to-origin scheme in the prior art, this application can reduce redundant paths and boundary risks in the return process, and improve the success rate, stability, and execution efficiency of the return-to-origin action in complex scenarios.

[0713] As can be seen from the above embodiments, the purpose of this application is to provide a full-coverage path planning method for window cleaning robots that addresses window type differences, thereby solving problems such as insufficient adaptability of existing window cleaning robot path planning schemes to different window types, unreasonable path mode selection, poor accuracy of resuming scanning after breakpoints, and a single strategy for returning to the origin. For example, Figure 4 This application provides a schematic diagram of the overall process for full-coverage path planning of a window cleaning robot based on window type differences, as illustrated in the embodiments of this application. Figure 4 As shown, the process includes:

[0714] Step 1: The window cleaning robot attaches to the target window and initializes the cleaning task and control parameters. Further, the window cleaning robot is controlled to acquire window feature information, including window width, frame status, and boundary combination. If it is a scenario for resuming cleaning after a breakpoint, the position information and breakpoint information of the window cleaning robot can also be acquired at this time.

[0715] Step 2: After obtaining the above information, the window cleaning robot determines the current task type and controls its own movement to perform edge probing and contact actions, identifying window shape differences, including wide / narrow windows, framed / frameless windows, left and right boundaries, etc. Further, based on the current task type, window feature information, and window shape differences, a path planning strategy is selected.

[0716] The current task type can include global cleaning tasks, edge cleaning tasks, zone cleaning tasks, and fixed-point cleaning tasks.

[0717] For example, Figure 5 A flowchart illustrating the selection of a window type recognition and path planning strategy is provided in this application embodiment, as follows: Figure 5 As shown, the process includes:

[0718] Step A: Control the window cleaning robot to move along the preset direction to the target boundary, complete the top edge probing and edge-fitting actions to measure and record the horizontal width of the window surface. Correspondingly, it is also necessary to record the border markings, such as the top border, bottom border, left border, and right border.

[0719] Step B: Determine if the horizontal width of the window is greater than the threshold of the fuselage dimensions. If yes, proceed to step C; otherwise, proceed to step D.

[0720] Step C: Select the path planning strategy as Z-shaped path and enter the path planning execution stage, where the main direction of movement of the Z-shaped path is lateral.

[0721] Step D: Further determine whether both the left and right boundaries exist. If both exist, select the N-type path planning strategy and perform attitude adjustment before executing the N-type path. The main motion direction of the N-type path is lateral. If the left and right boundaries are incomplete or without a frame, execute one of the following strategies: frameless processing, edge cleaning strategy, or return to the starting point strategy.

[0722] Step 3: Control the window cleaning robot to perform cleaning operations on the target window surface based on the selected path planning strategy. After the operation is completed, determine whether an edge-supplementary cleaning strategy is needed. If so, control the window cleaning robot to perform edge-supplementary cleaning based on the edge-supplementary cleaning strategy. If not, select a differentiated return-to-start strategy based on the path planning strategy and window surface feature information to end the cleaning task.

[0723] In this scenario, if a breakpoint resumption request is detected during the window cleaning robot's cleaning task, the robot will execute a breakpoint resume cleaning strategy. For example, Figure 6 A flowchart for breakpoint continuation cleaning and window matching determination is provided in the embodiments of this application, such as... Figure 6 As shown, the process includes:

[0724] Step 1: If a breakpoint recovery request is detected, determine whether the breakpoint recovery flag is valid. If yes, proceed to Step 2; otherwise, control the window cleaning robot to restart cleaning according to the new cleaning task.

[0725] It should be noted that the determination of whether the breakpoint recovery flag is valid is the same as the determination of whether the breakpoint continuation cleaning strategy can be executed. This application will not elaborate further here, but for details, please refer to the description of the above embodiments.

[0726] Step 2: Re-execute the edge probing and edge-fitting actions, re-measure the current window width, and read the breakpoint information. The breakpoint information may include the historical window width, the distance of the breakpoint from the left and top boundaries, the number of path switching, etc.

[0727] Furthermore, the current window width is compared with the breakpoint information to determine if it meets the requirements for same-window detection. If so, it is determined to be the same window, and the breakpoint continuation cleaning strategy is executed, i.e., cleaning continues from the path corresponding to the breakpoint. If not, it is determined to be a different window, and the breakpoint recovery flag can be cleared, and the strategy can be replanned according to a new global cleaning task or a zone cleaning task.

[0728] It should be noted that the criteria for determining classmates can be found in the description of the preset determination conditions in the above embodiments, and will not be repeated here.

[0729] It should also be noted that, during the breakpoint recovery process, breakpoint recovery can also rely on methods such as marker information (QR code or marker point) comparison, visual map matching, or user confirmation. This application embodiment does not impose specific limitations on this.

[0730] Based on the analysis of the above embodiments, it can be seen that this application has at least the following beneficial effects:

[0731] 1. Improve adaptability to different window types:

[0732] This application utilizes a comprehensive approach, incorporating window width, border presence status, boundary combination relationships, region attributes, and interruption information, to achieve finer-grained identification of the current window type and adaptively select different path planning strategies accordingly. Compared to existing technologies that primarily rely on fixed path templates, this application better adapts to different operational scenarios, including wide windows, narrow windows, framed windows, frameless windows, and partially partitioned windows, thereby improving the applicability of window cleaning robots under complex window conditions.

[0733] 2. Improve the completeness of window surface cleaning:

[0734] This application does not simply apply a uniform Z-shaped or N-shaped path to all windows. Instead, it first identifies differences in window types and then determines a more suitable main coverage path and supplementary cleaning method for the current window. Therefore, it can effectively reduce problems such as partial omissions, uncovered edges, and insufficient coverage caused by mismatch between the path pattern and the window type, thereby improving the integrity and reliability of full-coverage window cleaning.

[0735] 3. Reduce repeated wiping and unnecessary paths, improving work efficiency:

[0736] This application dynamically selects a path planning strategy based on the relationship between the window width and the robot's dimensions, as well as boundary combinations. This enables more efficient coverage paths under suitable window conditions and avoids blindly executing unreasonable path switching in unsuitable scenarios. This reduces repetitive wiping, ineffective turning, and redundant return trips, shortens the overall cleaning path length and operation time, and improves the cleaning efficiency and energy utilization efficiency of the window cleaning robot.

[0737] 4. Improve the accuracy and continuity of breakpoint re-erasing:

[0738] This application, when resuming from a breakpoint, does not rely solely on historical location or path for continuation scanning. Instead, it combines the current window width and window boundary features to determine whether the current window surface is the same piece of glass as the original breakpoint window surface. By introducing a same-window discrimination mechanism, it effectively avoids problems such as incorrect continuation scanning, misaligned recovery, or incompatibility of the original path caused by the user repositioning the device to another piece of glass, thereby improving the accuracy, continuity, and practical usability of breakpoint continuation scanning.

[0739] 5. Improve the stability and rationality of the return-to-starting-point process:

[0740] This application can select differentiated return-to-origin strategies based on different window types and path modes. For example, different return methods can be used for narrow windows, frameless windows, and different main path modes. Compared with the unified return-to-origin scheme in the prior art, this application can reduce redundant paths and boundary risks in the return process, and improve the success rate, stability, and execution efficiency of the return-to-origin action in complex window type scenarios.

[0741] 6. Improve security in frameless and complex boundary scenarios:

[0742] This application considers not only physical borders during path planning, but also frameless edges and boundary combinations, and dynamically adjusts the path pattern, edge-following logic, and return logic as needed. This avoids using conventional template paths in scenarios with frameless edges, incomplete boundaries, or mixed boundaries, thereby reducing the risks of falls, abnormal shutdowns, and operational risks caused by edge misjudgments, and improving the safety of the window cleaning robot in complex boundary scenarios.

[0743] 7. Improve the coordination between global cleaning tasks, zone cleaning tasks, and edge cleaning tasks:

[0744] This application integrates global cleaning tasks, zone cleaning tasks, edge-based supplementary cleaning, breakpoint recovery, and return to the starting point into a unified planning framework. This prevents the sub-processes from becoming isolated and enables them to make coordinated decisions based on the current window characteristics. This approach helps maintain consistency in path strategies across different tasks and ensures stable cleaning logic during task switching or abnormal recovery, thereby improving the overall coordination of the entire system control.

[0745] 8. Enhance user experience and product intelligence:

[0746] Because this application can automatically select appropriate cleaning paths, re-cleaning methods, and recovery strategies for different window types, users do not need to frequently intervene manually or reselect path modes, thus obtaining a more stable, efficient, and complete automatic cleaning experience. Therefore, this application helps to improve the intelligence level, environmental adaptability, and market competitiveness of window cleaning robot products.

[0747] In the foregoing embodiments, the control method for the window cleaning robot provided in this application has been described. To achieve the functions of the methods provided in the embodiments of this application, the window cleaning robot, as the executing entity, may include hardware structures and / or software modules, implementing the above functions in the form of hardware structures, software modules, or a combination of hardware structures and software modules. Whether a particular function is executed in the form of hardware structures, software modules, or a combination of hardware structures and software modules depends on the specific application and design constraints of the technical solution.

[0748] For example, Figure 7 This is a schematic diagram of the control device for a window cleaning robot provided in an embodiment of this application, as shown below. Figure 7 As shown, the control device 700 of the window cleaning robot includes:

[0749] The acquisition module 701 is used to acquire the surface features of the surface to be cleaned by the window cleaning robot. The surface features include surface width information, boundary type information, and boundary combination state.

[0750] The control module 702 is used to select the corresponding path planning strategy according to the surface characteristics, and control the window cleaning robot to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

[0751] Optionally, the control module 702 includes a selection unit for:

[0752] If the surface width information is determined to be greater than the preset width threshold, the path planning strategy is determined to be the first strategy, and the main travel direction of the cleaning path corresponding to the first strategy is horizontal.

[0753] If the surface width information is determined to be less than or equal to the preset width threshold, and the boundary combination state is that both the left and right boundaries have borders, the path planning strategy is determined to be the second strategy, and the main travel direction of the cleaning path corresponding to the second strategy is longitudinal.

[0754] Optionally, the control module 702 includes a control unit, which is used for:

[0755] When the first strategy is adopted, the window cleaning robot is controlled to move back and forth laterally from the starting position, and each time it reaches the first target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a lateral strip path covering the surface to be cleaned.

[0756] The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on a horizontal strip path.

[0757] Optionally, the control module 702 is also used for:

[0758] After the window cleaning robot finishes cleaning the surface to be cleaned based on the horizontal strip path, the robot is controlled to perform an edge cleaning strategy and return to the starting position.

[0759] Optionally, the control unit is used for:

[0760] When the second strategy is adopted, the window cleaning robot is controlled to move back and forth in the longitudinal direction, and each time it reaches the second target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a longitudinal strip path covering the surface to be cleaned;

[0761] The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on a longitudinal strip path.

[0762] Optionally, before controlling the window cleaning robot to move back and forth longitudinally, the control module 702 is also used to:

[0763] Control the window cleaning robot to perform a lateral edge probing and width measurement operation to obtain the available lateral width of the surface to be cleaned;

[0764] After obtaining the available horizontal width, the body posture of the window cleaning robot is adjusted;

[0765] If the robot body is found to be obstructed during the posture adjustment process, the robot body is controlled to retreat longitudinally by a preset safe distance, and the robot body posture is readjusted again so that the front side of the robot body contacts the boundary of the surface to be cleaned.

[0766] Based on the type of signal generated when the front side of the fuselage contacts the boundary, determine whether there is a restricted area on the surface to be cleaned;

[0767] After determining that there are restricted areas on the surface to be cleaned, the number of strips for the longitudinal strip path is determined based on the available lateral width.

[0768] Optionally, this selection unit is used for:

[0769] Obtain the current task type of the window cleaning robot, and select the corresponding path planning strategy based on the current task type and surface features.

[0770] Optionally, this selection unit is specifically used for:

[0771] If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that at least one of the left and right boundaries has a border, control the window cleaning robot to execute the edge cleaning strategy so that the window cleaning robot moves and cleans along the boundary of the surface to be cleaned.

[0772] If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that neither the left nor the right boundary has a border, or the surface width information is less than or equal to a preset threshold, the window cleaning robot is controlled to execute a frameless return-to-start strategy so that the window cleaning robot returns to the starting position.

[0773] Optionally, this selection unit is specifically used for:

[0774] If the current task type is determined to be a global cleaning task, and the boundary combination state is determined to be that both the left and right boundaries have borders, and the surface width information is greater than a preset threshold, the window cleaning robot is controlled to execute a full-coverage path strategy.

[0775] If the current task type is determined to be a global cleaning task, and the surface features are determined to meet at least one of the following conditions: the boundary combination state is that at least one of the left and right boundaries has no border, the surface width information is less than or equal to a preset threshold, or no surface width information is detected, then the window cleaning robot is controlled to execute a protection strategy. The protection strategy includes at least one of the following: a backtracking strategy, a return to the starting point strategy, or termination of the cleaning operation.

[0776] Optionally, this selection unit is specifically used for:

[0777] If the current task type is determined to be a local cleaning task, and the surface width information of the local area is determined to meet the preset conditions, the window cleaning robot is controlled to execute the first strategy.

[0778] If the current task type is determined to be a local cleaning task, and the surface width information of the local area does not meet the preset conditions, or the boundary combination state of the local area is that at least one of the left and right boundaries does not have a border, the window cleaning robot is controlled to execute a local strip coverage strategy so that the window cleaning robot can move back and forth to clean the local area based on the strip coverage path.

[0779] Optionally, this selection unit is specifically used for:

[0780] If the current task type is determined to be a zone cleaning task, the surface to be cleaned is divided into multiple target areas;

[0781] Select the corresponding path planning strategy based on the surface width and boundary type information of each target area.

[0782] Optionally, the control device 700 of the window cleaning robot further includes a first determining module, which is used for:

[0783] After the window cleaning robot has completed any of the tasks of zone cleaning, global cleaning, or edge cleaning, obtain the cleaning mode and boundary type information of the surface to be cleaned.

[0784] Based on cleaning mode and boundary type information, determine whether the window cleaning robot should perform an edge supplementary cleaning strategy.

[0785] Optionally, the control device 700 of the window cleaning robot further includes a second determining module, which is used for:

[0786] If it is determined that the window cleaning robot will perform a supplementary cleaning strategy along the edges, control the window cleaning robot to perform the following operations:

[0787] Determine the starting boundary for edge-to-edge cleaning based on boundary type information;

[0788] The window cleaning robot is controlled to move sequentially along the starting boundary in a first preset direction to perform supplementary cleaning on the edge areas of the surface to be cleaned;

[0789] Record the number of times the window cleaning robot cleans along the edges, and when the number of times it cleans along the edges reaches the preset number, control the window cleaning robot to execute the return-to-start strategy.

[0790] Optionally, the control unit 700 of the window cleaning robot also includes a supplementary cleaning module, which is used for:

[0791] Obtain the number of edge loops and / or edge entry conditions corresponding to the execution of the edge supplementary cleaning strategy;

[0792] The window cleaning robot is controlled to perform a supplementary cleaning strategy along the edges based on the number of edge loops and / or edge entry conditions.

[0793] Optionally, module 701 is used to obtain:

[0794] The window cleaning robot is controlled to move along a second preset direction to the third target boundary of the surface to be cleaned, and performs edge probing and edge-adhering actions to obtain the surface features of the surface to be cleaned by the window cleaning robot.

[0795] Optionally, the control device 700 of the window cleaning robot also includes a third determining module, which is used for:

[0796] During the cleaning operation of the window cleaning robot on the surface to be cleaned, if an interruption is detected, the breakpoint information is recorded. The breakpoint information includes at least the surface width information corresponding to the surface to be cleaned that the window cleaning robot was adsorbed at the time of the interruption.

[0797] When it is determined that the window cleaning robot has resumed cleaning operations, the current surface characteristics of the surface to be cleaned by the window cleaning robot are reacquired.

[0798] Based on the current surface features and breakpoint information, determine whether to execute the breakpoint continuation cleaning strategy.

[0799] Optionally, the breakpoint information also includes the boundary type information and boundary combination state of the surface to be cleaned by the window cleaning robot when the interruption occurs; the third determination module is specifically used for:

[0800] If the current surface features and breakpoint information meet preset judgment conditions, the window cleaning robot is controlled to execute a breakpoint continuation cleaning strategy; the preset judgment conditions include at least one of the following:

[0801] The difference between the current surface width information and the surface width information in the breakpoint information is less than the preset difference threshold;

[0802] The current surface width is greater than or equal to the preset minimum width threshold;

[0803] The boundary type information and boundary combination status are consistent with the current boundary type information and current boundary combination status in the current surface features.

[0804] Optionally, the control device 700 of the window cleaning robot also includes a fourth determining module, which is used for:

[0805] If the current surface features and breakpoint information do not meet the preset judgment conditions, the corresponding path planning strategy is selected based on the current surface features, and the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

[0806] Optionally, the breakpoint information also includes the position information of the window cleaning robot at the time of the interruption. The third determination module is specifically used for:

[0807] Starting with the location information, the window cleaning robot continues to execute the corresponding path planning strategy to perform cleaning operations on the surface to be cleaned.

[0808] Optionally, the control device 700 of the window cleaning robot also includes a correction module, which is used for:

[0809] During the cleaning operation of the window cleaning robot on the surface to be cleaned, the feedback information of various sensors in the window cleaning robot is detected;

[0810] When feedback indicates that the window cleaning robot has encountered an anomaly or boundary anomaly, the robot is controlled to correct its path planning strategy so that it can perform cleaning operations based on the corrected path planning strategy.

[0811] Optionally, this correction module is specifically used for:

[0812] Control the window cleaning robot to pause its cleaning operation;

[0813] After controlling the window cleaning robot to retreat a preset distance, it performs the cleaning operation again based on the path planning strategy;

[0814] After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy.

[0815] Control the window cleaning robot to execute the edge cleaning strategy.

[0816] Optionally, the control device 700 of the window cleaning robot further includes a fifth determining module, which is used for:

[0817] During the cleaning operation of the window cleaning robot on the surface to be cleaned, if the window cleaning robot encounters abnormal working conditions, the type of abnormality is determined.

[0818] Based on the type of exception, the window cleaning robot selects the corresponding processing strategy so that the window cleaning robot can perform the corresponding operation based on the processing strategy.

[0819] Optionally, this fifth determining module is specifically used for:

[0820] Control the window cleaning robot to perform local obstacle avoidance actions, and after obstacle avoidance, continue to perform cleaning operations based on the selected path planning strategy;

[0821] After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy.

[0822] Control the window cleaning robot to prevent it from continuing the breakpoint cleaning strategy;

[0823] Control the window cleaning robot to pause its cleaning operation;

[0824] Control the window cleaning robot to execute the return-to-start strategy.

[0825] Optionally, the control device 700 of the window cleaning robot also includes a selection control module, which is used for:

[0826] After completing the cleaning operation, the corresponding return-to-start strategy is selected based on the path planning strategy and / or surface characteristics, and the window cleaning robot is controlled to return to the starting position based on the return-to-start strategy.

[0827] Optionally, this selection control module is specifically used for:

[0828] When the path planning strategy is a strategy with the longitudinal direction as the main direction of travel, the window cleaning robot is controlled to return to the starting point along the longitudinal direction.

[0829] When the path planning strategy is a strategy with the lateral direction as the main travel direction, the general strategy for controlling the window cleaning robot is to return to the starting point.

[0830] When the surface features indicate that the boundary of the surface to be cleaned is incomplete, the window cleaning robot adopts a frameless return-to-start strategy.

[0831] When the path planning strategy is an edge cleaning strategy, the current boundary and starting edge information are determined based on the surface features. After adjusting the body posture of the window cleaning robot based on the current boundary and starting edge information, the window cleaning robot is controlled to adopt a general return-to-starting-point strategy.

[0832] It should be noted that the specific implementation principle and effect of the control device 700 of the above-mentioned window cleaning robot can be found in the relevant description and effect of the above embodiments, and will not be elaborated further here.

[0833] This application also provides a window cleaning robot. Figure 8 This is a schematic diagram of the structure of a window cleaning robot provided in an embodiment of this application, as shown below. Figure 8 As shown, the window cleaning robot may include: a processor 801 and a memory 802 communicatively connected to the processor 801; the memory 802 stores a computer program; the processor 801 executes the computer program stored in the memory 802, causing the processor 801 to perform the method described in any of the above embodiments.

[0834] The memory 802 and the processor 801 can be connected via bus 803.

[0835] This application also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, are used to implement the methods described in any of the foregoing embodiments of this application.

[0836] This application also provides a chip for executing instructions, which is used to perform the methods described in any of the foregoing embodiments executed by an electronic device as described in any of the foregoing embodiments of this application.

[0837] This application also provides a computer program product, which includes a computer program that, when executed by a processor, can implement the methods described in any of the foregoing embodiments executed by an electronic device as described in any of the foregoing embodiments of this application.

[0838] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0839] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0840] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0841] The integrated modules implemented as software functional modules described above can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application.

[0842] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0843] The memory may include high-speed random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0844] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0845] The aforementioned storage media can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as Static Random-Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage media can be any available medium accessible to general-purpose or special-purpose computers.

[0846] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. Both the processor and the storage medium can reside in application-specific integrated circuits (ASICs). Alternatively, the processor and storage medium can exist as discrete components in an electronic device or host device.

[0847] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0848] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0849] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0850] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0851] The above are merely specific implementations of the embodiments of this application, but the protection scope of the embodiments of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the embodiments of this application should be covered within the protection scope of the embodiments of this application.

Claims

1. A control method for a window cleaning robot, characterized in that, The method includes: The surface features of the surface to be cleaned by the window cleaning robot are obtained, including surface width information, boundary type information, and boundary combination state. Based on the surface features, a corresponding path planning strategy is selected, and the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

2. The method according to claim 1, characterized in that, The step of selecting the corresponding path planning strategy based on the surface features includes: If the surface width information is determined to be greater than a preset width threshold, the path planning strategy is determined to be the first strategy, and the main travel direction of the cleaning path corresponding to the first strategy is horizontal. If the surface width information is determined to be less than or equal to the preset width threshold, and the boundary combination state is that both the left and right boundaries have borders, the path planning strategy is determined to be the second strategy, and the main travel direction of the cleaning path corresponding to the second strategy is longitudinal.

3. The method according to claim 2, characterized in that, The process of controlling the window cleaning robot to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy includes: When the first strategy is adopted, the window cleaning robot is controlled to move back and forth laterally from the starting position, and each time it reaches the first target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a lateral strip path covering the surface to be cleaned; The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the horizontal strip path.

4. The method according to claim 3, characterized in that, The method further includes: After the window cleaning robot finishes cleaning the surface to be cleaned based on the horizontal strip path, the robot is controlled to perform an edge cleaning strategy and return to the starting position.

5. The method according to claim 2, characterized in that, The process of controlling the window cleaning robot to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy includes: When the second strategy is adopted, the window cleaning robot is controlled to move back and forth in the longitudinal direction, and each time it reaches the second target boundary, the window cleaning robot is controlled to perform a turning movement operation to form a longitudinal strip path covering the surface to be cleaned; The window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the longitudinal strip path.

6. The method according to claim 5, characterized in that, Before controlling the window cleaning robot to move back and forth longitudinally, the method further includes: The window cleaning robot is controlled to perform a lateral edge probing and width measurement operation to obtain the available lateral width of the surface to be cleaned; After obtaining the available lateral width, the body posture of the window cleaning robot is adjusted; If the robot body is detected to be obstructed during the posture adjustment process, the robot body is controlled to retreat longitudinally by a preset safe distance, and the robot body posture is readjusted again so that the front side of the robot body contacts the boundary of the surface to be cleaned. Based on the type of signal generated when the front side of the fuselage contacts the boundary, it is determined whether there is a restricted area on the surface to be cleaned; After determining that the restricted area exists on the surface to be cleaned, the number of strips in the longitudinal strip path is determined based on the available lateral width.

7. The method according to claim 1, characterized in that, The step of selecting the corresponding path planning strategy based on the surface features includes: Obtain the current task type of the window cleaning robot, and select the corresponding path planning strategy based on the current task type and the surface features.

8. The method according to claim 7, characterized in that, The method of selecting the corresponding path planning strategy based on the current task type and the surface features includes: If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that at least one of the left and right boundaries has a border, the window cleaning robot is controlled to execute the edge cleaning strategy so that the window cleaning robot moves and cleans along the boundary of the surface to be cleaned. If the current task type is determined to be an edge cleaning task, and the boundary combination state is determined to be that neither the left nor the right boundary has a border, or the surface width information is less than or equal to a preset threshold, the window cleaning robot is controlled to execute a frameless return-to-start strategy so that the window cleaning robot returns to the starting position.

9. The method according to claim 7, characterized in that, The method of selecting the corresponding path planning strategy based on the current task type and the surface features includes: If the current task type is determined to be a global cleaning task, and the boundary combination state is determined to be that both the left and right boundaries have borders, and the surface width information is greater than a preset threshold, the window cleaning robot is controlled to execute a full-coverage path strategy. If the current task type is determined to be a global cleaning task, and the surface feature is determined to meet at least one of the following conditions: the boundary combination state is that at least one of the left and right boundaries has no border, the surface width information is less than or equal to a preset threshold, or the surface width information is not detected, then the window cleaning robot is controlled to execute a protection strategy. The protection strategy includes at least one of the following: a backtracking strategy, a return to the starting point strategy, or termination of the cleaning operation.

10. The method according to claim 7, characterized in that, The method of selecting the corresponding path planning strategy based on the current task type and the surface features includes: If the current task type is determined to be a local cleaning task, and the surface width information of the local area meets the preset conditions, the window cleaning robot is controlled to execute the first strategy. If the current task type is determined to be a local cleaning task, and the surface width information of the local area does not meet the preset conditions, or the boundary combination state of the local area is that at least one of the left and right boundaries does not have a border, the window cleaning robot is controlled to execute a local strip coverage strategy so that the window cleaning robot can perform back-and-forth cleaning based on the strip coverage path in the local area.

11. The method according to claim 7, characterized in that, The method of selecting the corresponding path planning strategy based on the current task type and the surface features includes: If the current task type is determined to be a zone cleaning task, the surface to be cleaned is divided into multiple target areas; Select the corresponding path planning strategy based on the surface width and boundary type information of each target area.

12. The method according to claim 1, characterized in that, The method further includes: After the window cleaning robot has completed any of the following tasks: zone cleaning, global cleaning, or edge cleaning, the cleaning mode and boundary type information corresponding to the surface to be cleaned is obtained. Based on the cleaning mode and the boundary type information, it is determined whether the window cleaning robot should perform an edge-supplementary cleaning strategy.

13. The method according to claim 12, characterized in that, The method further includes: If it is determined that the window cleaning robot is executing the edge-sweeping supplementary cleaning strategy, control the window cleaning robot to perform the following operations: The starting boundary for edge-to-edge cleaning is determined based on the boundary type information. The window cleaning robot is controlled to move sequentially along the starting boundary in a first preset direction to perform supplementary cleaning on the edge area of ​​the surface to be cleaned; The number of times the window cleaning robot performs supplementary cleaning along the edges is recorded, and when the number of times reaches the preset number of times along the edges, the window cleaning robot is controlled to execute a return-to-starting-point strategy.

14. The method according to claim 12, characterized in that, The method further includes: Obtain the number of edge loops and / or edge entry conditions corresponding to the execution of the edge supplementary cleaning strategy; The window cleaning robot is controlled to execute the edge supplementary cleaning strategy based on the number of edge loops and / or the edge entry conditions.

15. The method according to claim 1, characterized in that, The process of acquiring the surface features of the surface to be cleaned by the window cleaning robot includes: The window cleaning robot is controlled to move along a second preset direction to the third target boundary of the surface to be cleaned, and performs edge probing and edge-adhering actions to obtain the surface features of the surface to be cleaned by the window cleaning robot.

16. The method according to claim 1, characterized in that, The method further includes: During the cleaning operation performed by the window cleaning robot on the surface to be cleaned, if an interruption is detected, the breakpoint information is recorded. The breakpoint information includes at least the surface width information corresponding to the surface to be cleaned by the window cleaning robot at the time of the interruption. When it is determined that the window cleaning robot has resumed cleaning operations, the current surface characteristics of the surface to be cleaned by the window cleaning robot are reacquired. Based on the current surface features and the breakpoint information, determine whether to execute the breakpoint continuation cleaning strategy.

17. The method according to claim 16, characterized in that, The breakpoint information also includes the boundary type information and boundary combination state corresponding to the surface to be cleaned by the window cleaning robot when the interruption occurs; the step of determining whether to execute the breakpoint continuation cleaning strategy based on the current surface features and the breakpoint information includes: If the current surface features and the breakpoint information meet a preset judgment condition, the window cleaning robot is controlled to execute a breakpoint continuation cleaning strategy; the preset judgment condition includes at least one of the following: The difference between the current surface width information and the surface width information in the breakpoint information is less than a preset difference threshold; The current surface width information is greater than or equal to a preset minimum width threshold; The boundary type information and boundary combination state are consistent with the current boundary type information and current boundary combination state in the current surface feature.

18. The method according to claim 17, characterized in that, The method further includes: If the current surface features and the breakpoint information do not meet the preset judgment conditions, a corresponding path planning strategy is selected based on the current surface features, and the window cleaning robot is controlled to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

19. The method according to claim 17, characterized in that, The breakpoint information also includes the position information of the window cleaning robot at the time of the interruption, and the step of controlling the window cleaning robot to execute the breakpoint continuation cleaning strategy includes: Starting from the location information, the window cleaning robot continues to execute the corresponding path planning strategy to perform cleaning operations on the surface to be cleaned.

20. The method according to claim 1, characterized in that, The method further includes: During the cleaning operation performed by the window cleaning robot on the surface to be cleaned, feedback information from various sensors in the window cleaning robot is detected. If the feedback information indicates that the window cleaning robot has encountered an anomaly or a boundary anomaly, the window cleaning robot is controlled to correct the path planning strategy so that the window cleaning robot performs cleaning operations based on the corrected path planning strategy.

21. The method according to claim 20, characterized in that, The window cleaning robot performs cleaning operations based on a modified path planning strategy, including any of the following: The window cleaning robot is controlled to pause the cleaning operation. After controlling the window cleaning robot to retreat a preset distance, the cleaning operation is performed again based on the path planning strategy; After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy. Control the window cleaning robot to execute an edge cleaning strategy.

22. The method according to claim 1, characterized in that, The method further includes: During the cleaning operation of the window cleaning robot on the surface to be cleaned, if the window cleaning robot encounters an abnormal working condition, the abnormal type of the abnormal working condition shall be determined. Based on the type of anomaly, the window cleaning robot is controlled to select the corresponding processing strategy, so that the window cleaning robot performs the corresponding operation based on the processing strategy.

23. The method according to claim 22, characterized in that, The window cleaning robot performs corresponding operations based on the processing strategy, including any one of the following: The window cleaning robot is controlled to perform local obstacle avoidance actions, and after obstacle avoidance, it continues to perform cleaning operations based on the selected path planning strategy; After the window cleaning robot switches paths again, it performs the cleaning operation again based on the path planning strategy. The window cleaning robot is prevented from continuing the breakpoint-resume cleaning strategy. The window cleaning robot is controlled to pause the cleaning operation. Control the window cleaning robot to execute the return-to-start strategy.

24. The method according to claim 1, characterized in that, The method further includes: After completing the cleaning operation, the window cleaning robot selects a corresponding return-to-start strategy based on the path planning strategy and / or the surface features, and controls the window cleaning robot to return to the starting position based on the return-to-start strategy.

25. The method according to claim 24, characterized in that, The step of selecting the corresponding return-to-start strategy based on the path planning strategy and / or the surface features includes: When the path planning strategy is a strategy with the longitudinal direction as the main travel direction, the window cleaning robot is controlled to adopt a strategy of returning to the starting point in the longitudinal direction. When the path planning strategy is a strategy with the lateral direction as the main travel direction, the window cleaning robot is controlled to return to the starting point using a general strategy. When the surface features indicate that the boundary of the surface to be cleaned is incomplete, the window cleaning robot is controlled to adopt a frameless return-to-start strategy. When the path planning strategy is an edge cleaning strategy, the current boundary and starting edge information are determined based on the surface features. After adjusting the body posture of the window cleaning robot based on the current boundary and starting edge information, the window cleaning robot is controlled to adopt a general return-to-starting-point strategy.

26. A control device for a window cleaning robot, characterized in that, The device includes: The acquisition module is used to acquire the surface features of the surface to be cleaned by the window cleaning robot, the surface features including surface width information, boundary type information and boundary combination state; The control module is used to select a corresponding path planning strategy based on the surface features, and control the window cleaning robot to perform cleaning operations on the surface to be cleaned based on the selected path planning strategy.

27. A window cleaning robot, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-25.

28. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-25.

29. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method as described in any one of claims 1-25.

Images