Surface laying object cleaning method and device based on mobile cleaning robot

Abstract

The invention discloses a surface laying object cleaning method based on a mobile cleaning robot, and the method comprises the steps: obtaining the boundary information of a to-be-cleaned surface laying object and the type of the surface laying object, carrying out the low-risk boundary detection of the boundary of the to-be-cleaned surface laying object, so as to determine the low-risk boundary information, the low-risk boundary information is used for representing boundary information facilitating the movement of the mobile cleaning robot to the upper surface of the surface laying object, and determining an upward movement strategy of the mobile cleaning robot according to the obtained type of the surface laying object, according to the upward moving strategy, the surface laying object is not pushed in the upward moving process of the mobile cleaning robot moving to the upper surface of the surface laying object, and the mobile cleaning robot moves to the upper surface of the surface laying object to be cleaned according to the determined upward moving strategy at the low-risk boundary position determined based on the low-risk boundary information; and cleaning is conducted on the upper surface of the to-be-cleaned surface laying object according to a cleaning path planned by the surface laying object. According to the invention, the reliability and efficiency of cleaning the surface laying object are improved.

Description

Technical Field

[0001] The present invention relates to the field of mobile robot cleaning, and in particular to a surface covering cleaning method based on a mobile cleaning robot. Background Art

[0002] In recent years, mobile cleaning robots have gradually entered the end-use market, boasting increasingly diverse functions. However, their cleaning performance on surface coverings such as carpets and floor mats is often suboptimal. For example, mobile cleaning robots frequently become stuck on, push, or even crawl under surface coverings, resulting in a poor user experience.

[0003] Existing surface cleaning methods often struggle to perform in real-world scenarios. Some methods, limited by the surface's boundary, simply forcibly move up to the surface and clean it, ignoring various factors influencing the robot's upward movement. Others determine whether to clean the surface based on the height of the surface's lint, ignoring the robot's ability to successfully reach the surface. Summary of the Invention

[0004] The present invention provides a surface covering cleaning method based on a mobile cleaning robot, so as to improve the reliability of the mobile cleaning robot moving to the upper surface of the surface covering.

[0005] A first aspect of the present invention provides a surface covering cleaning method based on a mobile cleaning robot, the method comprising: on the mobile cleaning robot side,

[0006] Get the boundary information and surface covering type of the surface to be cleaned,

[0007] Low-risk boundary detection is performed on the boundary of the surface covering to be cleaned to determine low-risk boundary information, where the low-risk boundary information is used to represent boundary information that is conducive to the mobile cleaning robot moving to the upper surface of the surface covering.

[0008] Determine an upward movement strategy of the mobile robot according to the acquired surface covering category, wherein the upward movement strategy ensures that the surface covering itself is not pushed during the upward movement of the mobile cleaning robot to the upper surface of the surface covering.

[0009] At the low-risk boundary position determined based on the low-risk boundary information, according to the determined upward movement strategy, move to the upper surface of the surface to be cleaned.

[0010] On the surface of the surface to be cleaned, clean according to the cleaning path planned on the surface.

[0011] Preferably, determining the upward movement strategy of the mobile robot according to the acquired surface covering category includes:

[0012] Determine the strength of adhesion between the surface covering and the supporting surface and the height of the surface covering according to the type of surface covering.

[0013] According to the strength of the adhesion between the surface covering and the supporting surface and the height of the surface covering, the corresponding upward movement strategy is determined. The upward movement strategy is used to determine the contact state between the walking wheels of the mobile cleaning robot and the boundary of the surface covering during the upward movement.

[0014] Preferably, the mobile cleaning robot has a driving wheel and a driven wheel, wherein the driven wheel is located close to the front of the mobile cleaning robot and the driving wheel is located away from the front of the mobile cleaning robot. The upward movement strategy is determined according to the adhesion between the surface covering and the supporting surface and the height of the surface covering, including:

[0015] In the case of strong adhesion between the surface covering and the bearing surface,

[0016] If the height of the surface covering is not greater than the set first height threshold, the first method is adopted.

[0017] If the height of the surface covering is greater than the set first height threshold, the second method is used.

[0018] In cases where the adhesion between the surface covering and the bearing surface is weak,

[0019] If the height of the surface covering is not greater than the set second height threshold, the third method is adopted.

[0020] If the height of the surface covering is greater than the set second height threshold, the fourth method is adopted.

[0021] in,

[0022] The first method is: during the upward movement, each driving wheel of the mobile cleaning robot simultaneously contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering, and each driven wheel moves upward to the upper surface of the surface covering before the driving wheel contacts the boundary of the surface covering.

[0023] The second method is: during the upward movement, each driving wheel of the mobile cleaning robot sequentially contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering, and each driven wheel contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering before the driving wheel.

[0024] The third method is: during the upward movement, each driving wheel of the mobile cleaning robot simultaneously contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering, and each driven wheel moves upward to the upper surface of the surface covering after the driving wheel contacts the boundary of the surface covering.

[0025] The fourth method is: during the upward movement, each driving wheel of the mobile cleaning robot contacts the boundary of the surface covering one by one and moves up to the upper surface of the surface covering, and each driven wheel contacts the boundary of the surface covering and moves up to the upper surface of the surface covering later.

[0026] Preferably, in a case where the mobile cleaning robot has only driving wheels but no driven wheels, the upward movement strategy is used to determine the angle relationship between the moving direction of the mobile cleaning robot and the boundary of the surface covering during the upward movement.

[0027] Preferably, in the case where the mobile cleaning robot has only driving wheels but no driven wheels, the upward movement strategy is used to determine the angle relationship between the moving direction of the mobile cleaning robot and the boundary of the surface covering during the upward movement process.

[0028] The determining of the corresponding upward movement strategy according to the strength of the adhesion between the surface covering and the bearing surface and the height of the surface covering includes:

[0029] In the case of strong adhesion between the surface covering and the bearing surface,

[0030] If the height of the surface covering is not greater than the set first height threshold, the upward method is adopted.

[0031] If the height of the surface covering is greater than the set first height threshold, the inclined upward method is adopted.

[0032] In cases where the adhesion between the surface covering and the bearing surface is weak,

[0033] If the height of the surface covering is not greater than the set second height threshold, the reverse-up method is adopted.

[0034] If the height of the surface covering is greater than the set second height threshold, the reverse oblique method is adopted. Preferably, the first method is the straight method, the second method is the oblique method, the third method is the reverse straight method, and the fourth method is the reverse oblique method.

[0035] in,

[0036] The above method is: the moving direction of the mobile cleaning robot is consistent with the normal direction of the front of the mobile cleaning robot, and the moving direction is perpendicular to the boundary of the surface covering.

[0037] The oblique upward method is: the moving direction of the mobile cleaning robot is consistent with the normal direction of the front of the mobile cleaning robot, and the moving direction forms an angle with the boundary of the surface covering,

[0038] The reverse and correct method is: the moving direction of the mobile cleaning robot is opposite to the normal direction of the front of the mobile cleaning robot, and the moving direction is perpendicular to the boundary of the surface covering.

[0039] The inverted oblique upward method is that the moving direction of the mobile cleaning robot is opposite to the normal direction of the front of the mobile cleaning robot, and the moving direction forms an angle with the boundary of the surface covering.

[0040] Preferably, the angle is between 5° and 85°.

[0041] Preferably, the step of obtaining boundary information and surface covering category of the surface covering to be cleaned includes:

[0042] The mobile cleaning robot obtains current image data while performing a cleaning task, and obtains regional information of the surface covering to be cleaned and the type of the surface covering from the current image data;

[0043] Preferably, the cleaning of the upper surface of the surface covering to be cleaned according to the cleaning path planned for the surface covering includes:

[0044] Surface covering map data is acquired from the map data, and based on the surface covering map data, a bow-shaped cleaning path is planned with the long side of the surface covering so as to perform cleaning according to the cleaning path.

[0045] Preferably, the angle is 45°.

[0046] Preferably, the step of obtaining the area information of the surface covering to be cleaned and the type of the surface covering from the current image data includes:

[0047] The trained deep learning segmentation network model is used to identify the current image data and obtain the surface covering area information in the current image data.

[0048] Using the trained deep neural network model, the surface covering area in the current image data is identified to obtain the surface covering category;

[0049] Preferably, determining low-risk boundary information based on the acquired surface covering area information includes:

[0050] Extracting the boundary information of the surface covering from the acquired surface covering area information,

[0051] For the extracted surface covering boundary, the boundary gradient in more than one sliding window is calculated to obtain more than one boundary gradient.

[0052] For each boundary gradient, if the boundary gradient is not greater than a set boundary gradient threshold, the boundary in the sliding window corresponding to the boundary gradient is determined to be a low-risk boundary.

[0053] Preferably, there are multiple sliding windows, each of which has the same size, and each of which is continuously adjacent to the boundary.

[0054] Preferably, the boundary gradient is calculated as follows:

[0055] The ratio between the one-dimensional position change and the other-dimensional position change of the extracted surface covering boundary within a sliding window is calculated to obtain the boundary gradient within the sliding window.

[0056] Preferably, the second aspect of the present application provides a surface covering cleaning device based on a mobile cleaning robot, the device comprising a memory and a processor, the memory storing a computer program, the processor being configured to execute the computer program to perform any of the steps of the surface covering cleaning method based on a mobile cleaning robot.

[0057] Preferably, the third aspect of the present application provides a mobile cleaning robot, comprising a surface covering cleaning device based on the mobile cleaning robot.

[0058] The surface covering cleaning method based on the mobile cleaning robot provided in the embodiment of the present application can identify uneven risk points such as warped edges and protrusions on the surface covering boundaries through low-risk boundary detection, and can effectively avoid the cleaning risk problems caused by uneven risk boundaries; at the low-risk boundary position, by adopting corresponding upward movement strategies for different surface covering categories, the mobile cleaning robot can stably move up to the surface covering area for cleaning, which can not only avoid the mobile cleaning robot from pushing the surface covering, but also improve the reliability and efficiency of surface covering cleaning. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a flow chart of a surface covering cleaning method based on a mobile cleaning robot according to an embodiment of the present application.

[0060] Figure 2 A schematic diagram of a process for cleaning surface coverings according to an embodiment of the present application.

[0061] Figure 3 A schematic diagram of a visual image and a surface covering segmented based on the visual image according to an embodiment of the present application.

[0062] Figure 4a The embodiment of this application is based on Figure 3 A schematic diagram of projecting carpet boundary information extracted from a segmented surface covering image into a two-dimensional space coordinate system.

[0063] Figure 4b A schematic diagram of calculating boundary gradients according to an embodiment of the present application.

[0064] Figure 4c Another schematic diagram of projecting the extracted carpet boundary according to an embodiment of the present application.

[0065] Figure 5 A schematic diagram of the upward shift strategy in an embodiment of the present application.

[0066] Figure 6 This is a schematic diagram of a surface covering cleaning device based on a mobile cleaning robot according to an embodiment of the present application.

[0067] Figure 7 This is another schematic diagram of the surface covering cleaning device based on the mobile cleaning robot according to an embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical means and advantages of this application more clear, this application is further described in detail below with reference to the accompanying drawings.

[0069] In an embodiment of the present application, at a low-risk boundary position, the mobile cleaning robot moves from the supporting surface for supporting the surface covering body to the upper surface of the surface covering with an upward movement strategy corresponding to the category of the surface covering, so that the mobile cleaning robot can reliably clean the upper surface of the surface covering.

[0070] It should be understood that in the embodiments of the present application, the surface covering can be a carpet or mat on the floor in an indoor environment, or a covering on any plane, for example, a specific covering on a display surface. The lower surface of the surface covering is in contact with a bearing surface for bearing the main body of the surface covering, and the upper surface of the surface covering is the area to be cleaned. For surface coverings with a certain thickness, this results in a height difference between the bearing surface and the upper surface of the covering. The process of the mobile cleaning robot moving from the bearing surface to the upper surface of the covering is a process of moving from a low place to a high place, which is referred to as an upward movement process in the present application.

[0071] See also Figure 1 As shown, Figure 1 This is a flow chart of a surface cleaning method based on a mobile cleaning robot according to an embodiment of the present application. The method comprises: on the mobile cleaning robot side,

[0072] Step 101: Obtain boundary information and surface covering type of the surface covering to be cleaned.

[0073] As an example, the user can input the boundary information and surface covering category of the surface covering to be cleaned through the user interaction interface. For example, the boundary position information of the surface covering to be cleaned can be marked in the map data, and the surface covering category of the surface covering to be cleaned can be selected through the user interaction interface.

[0074] As another example, the mobile cleaning robot obtains current image data while performing a cleaning task, and obtains area information of the surface covering to be cleaned and the category of the surface covering from the current image data.

[0075] Step 102 is to perform low-risk boundary detection on the boundary of the surface covering to be cleaned to determine low-risk boundary information. The low-risk boundary information is used to represent boundary information that is unlikely to cause the mobile cleaning robot to be unable to move to the upper surface of the surface covering, that is, boundary information that is conducive to the mobile cleaning robot being unable to move to the upper surface of the surface covering.

[0076] In this step, low-risk boundary detection is used to detect whether there are curling edges, protrusions, etc. on the boundary of the surface covering that prevent the mobile cleaning robot from moving to the boundary of the surface covering, for example, an uneven boundary.

[0077] As an example, low-risk boundary detection can be performed using boundary gradients. For example, boundary gradients within one or more sliding windows are calculated to obtain one or more boundary gradients. For each boundary gradient, if the boundary gradient is not greater than a set boundary gradient threshold, the boundary within the sliding window corresponding to the boundary gradient is determined to be a low-risk boundary.

[0078] As an example, the boundary gradient is calculated as follows:

[0079] The ratio of the position change in one dimension to the position change in another dimension of the extracted surface covering boundary within a sliding window is calculated to obtain the boundary gradient within the sliding window. The sliding window is used to divide the projected surface covering boundary into a number of line segments, which is equivalent to fitting the projected surface covering boundary with a number of line segments to facilitate the calculation of the boundary gradient.

[0080] For example, in the image coordinate system, the ratio between the one-dimensional pixel position change and the other-dimensional pixel position change of the extracted surface covering boundary within a sliding window is calculated to obtain the boundary gradient within the sliding window.

[0081] For another example, the extracted surface covering boundary is projected into a two-dimensional coordinate system, and the two-dimensional coordinate system is used to correspond the image pixel position information of the surface covering boundary to the two-dimensional coordinate position information.

[0082] In a two-dimensional coordinate system, the ratio between the one-dimensional coordinate change and the other-dimensional coordinate change of the projected surface covering boundary within a sliding window is calculated to obtain the boundary gradient within the sliding window.

[0083] Step 103: Determine an upward movement strategy for the mobile cleaning robot based on the acquired surface covering category. The upward movement strategy ensures that the surface covering itself is not pushed when the mobile cleaning robot moves to the surface covering.

[0084] As an example, according to the type of surface covering, the strength of adhesion between the surface covering and the supporting surface and the height of the surface covering are determined, and according to the strength of adhesion between the surface covering and the supporting surface and the height of the surface covering, the corresponding upward movement strategy is determined. The upward movement strategy is used to determine the angle relationship between the moving direction of the mobile cleaning robot and the boundary of the surface covering during the upward movement process, and / or to determine the contact state between the walking wheels of the mobile cleaning robot and the boundary of the surface covering during the upward movement process.

[0085] As an example, the mobile cleaning robot has a driving wheel and a driven wheel, for example, a pair of driving wheels and a pair of driven wheels, the driving wheels are evenly distributed on the chassis of the mobile cleaning robot, the driven wheels are located close to the front face of the mobile cleaning robot, and the driving wheels are located away from the front face of the mobile cleaning robot, that is, the distance from the driven wheels to the front face is smaller than the distance from the driving wheels to the front face.

[0086] In the case of strong adhesion between the surface covering and the bearing surface,

[0087] If the height of the surface covering is not greater than the set first height threshold, the first method is adopted.

[0088] If the height of the surface covering is greater than the set first height threshold, the second method is used.

[0089] In cases where the adhesion between the surface covering and the bearing surface is weak,

[0090] If the height of the surface covering is not greater than the set second height threshold, the third method is adopted.

[0091] If the height of the surface covering is greater than the set second height threshold, the fourth method is adopted.

[0092] in,

[0093] The first method is: during the upward movement, each driving wheel of the mobile cleaning robot simultaneously contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering, and each driven wheel moves upward to the upper surface of the surface covering before the driving wheel contacts the boundary of the surface covering. In this way, since the driven wheel contacts the boundary of the surface covering before the driving wheel, it is possible to utilize the sensor installed on the front face for detection, and the strong adhesion of the surface covering can be utilized without fear of pushing the surface covering. By simultaneously contacting the boundary of the surface covering, it is beneficial to improve the efficiency of the cleaning robot in moving upward to low-height surface coverings.

[0094] The second method is: during the upward movement, the driving wheels of the mobile cleaning robot sequentially contact the boundary of the surface covering and move upward to the upper surface of the surface covering, and the driven wheels contact the boundary of the surface covering before the driving wheels and move upward to the upper surface of the surface covering. In this way, by sequentially contacting the boundary of the surface covering, the difficulty of the cleaning robot moving upward to the high surface covering is reduced, the power of the driving motor used to drive the driving wheels is avoided to increase, and the crushing pressure on the surface covering is reduced.

[0095] The third method is: during the upward movement, the driving wheels of the mobile cleaning robot simultaneously contact the boundary of the surface covering and move upward to the upper surface of the surface covering, and the driven wheels move upward to the upper surface of the surface covering after the driven wheels contact the boundary of the surface covering. In this way, the driving wheels contact the boundary of the surface covering before the driven wheels, so that the driving wheels can actively form a direct crushing force on the surface covering, thereby avoiding pushing the surface covering with weak adhesion during the upward movement. By simultaneously contacting the boundary of the surface covering, it is beneficial to improve the efficiency of the cleaning robot in moving up to low-height surface coverings, and to increase the crushing force formed by the driving wheels on the surface covering, thereby reducing the possibility of the surface covering being pushed.

[0096] A fourth method is as follows: during the upward movement process, each driving wheel of the mobile cleaning robot sequentially contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering. Each driven wheel then contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering after the driving wheel. This sequential contact with the boundary of the surface covering helps to simultaneously reduce the difficulty of the cleaning robot ascending to a high-altitude surface covering, thereby avoiding an increase in drive motor power and reducing the likelihood of the surface covering being pushed. It should be understood that the term "simultaneously" is used in contrast to "sequentially"; "simultaneously" can be understood as synchronous, while "sequentially" can be understood as asynchronous.

[0097] The first method is a straight-up method, the second method is an oblique-up method, the third method is an inverted straight-up method, and the fourth method is an inverted oblique-up method.

[0098] In the case where the mobile cleaning robot has only driving wheels and no driven wheels, for example, a pair of driving wheels,

[0099] In the case of strong adhesion between the surface covering and the bearing surface,

[0100] If the height of the surface covering is not greater than the set first height threshold, the upward method is adopted.

[0101] If the height of the surface covering is greater than the set first height threshold, the inclined upward method is adopted.

[0102] In cases where the adhesion between the surface covering and the bearing surface is weak,

[0103] If the height of the surface covering is not greater than the set second height threshold, the reverse-up method is adopted.

[0104] If the height of the surface covering is greater than the set second height threshold, the reverse oblique upward method is adopted. Steps 102 and 103 may not be in a strict order and may be performed in parallel.

[0105] Step 104: determine the low-risk boundary position based on the low-risk boundary information, and move to the upper surface of the surface to be cleaned at the low-risk boundary position according to the determined upward movement strategy.

[0106] Step 105 : Cleaning the upper surface of the surface covering to be cleaned according to the cleaning path planned for the surface covering.

[0107] As an example, a surface covering is treated as an independent area, and a bow-shaped cleaning path is planned according to the long side of the surface covering's upper surface dimensions. This embodiment of the application performs low-risk boundary detection on the boundaries of the surface covering to be cleaned, thereby preventing the mobile cleaning robot from moving from the boundary of a surface covering with curled edges, protrusions, etc. to the upper surface of the surface covering. By moving the robot upward to the upper surface of the surface covering using an upward movement strategy that matches the surface covering's type, the robot avoids the surface covering from being pushed, thereby improving the efficiency and reliability of cleaning the upper surface of the surface covering.

[0108] To facilitate understanding of this application, the following description uses a sweeper cleaning a carpet as an example. In this example, the side brush is mounted on the front of the sweeper, and the sweeper's travel wheels include drive wheels but not driven wheels. It should be understood that this embodiment is not limited to the specific structure of a sweeper or carpets themselves, but may also include other surface coverings such as carpets and floor mats.

[0109] See also Figure 2 As shown, Figure 2 This is a schematic diagram of a process for carpet cleaning according to an embodiment of the present application. The method includes:

[0110] Step 201: The mobile cleaning robot obtains current image data while performing a cleaning task, and obtains the carpet area location information and carpet category of the carpet to be cleaned from the current image data.

[0111] As an example, a trained deep learning segmentation network model such as HRNet-18 is used to identify the current visual image. The deep learning segmentation network model is an image semantic segmentation model. By training the deep learning segmentation network model with different types of carpet visual images as samples, the carpet image contained in the image can be segmented from the visual image. For example, see Figure 3 As shown, Figure 3 This diagram illustrates carpet segmentation based on a visual image. The top half of the image shows a visual image captured by the robot, while the bottom half shows the image segmented by the deep learning segmentation network model. The gray area represents the segmented carpet image. When a carpet image is detected in the visual image, positioning is triggered, and the carpet area's location information is retrieved from the map data.

[0112] As an example, when a carpet image is identified in a visual image, the trained deep neural network can be used to detect the carpet category of the current visual image. Generally, carpet categories based on their texture and production process include: silica algae mud carpet, plush carpet, velvet carpet, Saxony carpet, strong twist carpet, long pile carpet, flat loop pile carpet, Thai yuan carpet, and cut loop pile carpet. In this embodiment, the carpet cleaning process is primarily affected by the adhesion between the carpet and the floor, as well as the carpet's thickness. Lower adhesion makes it more difficult for the sweeper to move from the floor to the carpet's upper surface. Thicker carpets also increase the likelihood of difficulty in moving the carpet up. These difficulties in moving the carpet up are all considered difficulties in moving the carpet up. Therefore, in this embodiment, carpet category detection can be distinguished by silica algae mud carpet, thick pile carpet, looped silk carpet, and thin pile carpet. Silica algae mud carpet has the strongest adhesion to the floor, while thin pile carpet has weaker adhesion. Thick pile carpet and looped silk carpet have adhesion levels between silica algae mud carpet and thin pile carpet, respectively. The deep neural network can be trained using carpet visual image samples so that the deep neural network has the ability to detect carpet categories based on carpet visual images.

[0113] Step 202 : performing low-risk carpet boundary detection on the carpet boundary of the carpet to be cleaned to obtain low-risk carpet boundary information.

[0114] After obtaining the carpet area location information, the carpet area and the carpet boundary line can be determined based on the carpet area location information; alternatively, the discrete boundary points of the carpet can be obtained through the radar data reflected by the radar sensor, and then the carpet boundary line can be fitted based on the discrete boundary points. In addition, the radar data reflected by the radar sensor can also be used to identify the category of the carpet to verify the carpet category detected in step 201.

[0115] In this step, low-risk carpet boundary information refers to boundary information that is unlikely to cause the sweeper to be unable to move to the upper surface of the carpet, including the boundary position information of the carpet that can be moved, for example, the edge position information of the carpet boundary that does not have curling or raised areas and is conducive to the sweeper moving to the upper surface of the carpet.

[0116] To obtain low-risk carpet boundary information, the segmented carpet boundary image can be detected in the following way:

[0117] The boundary line information of the carpet is extracted from the segmented carpet image by a line extraction operator such as the Candy operator, and projected into a two-dimensional coordinate system, such as a Cartesian coordinate system, to obtain a boundary information set R, wherein the Cartesian coordinate system corresponds to the image coordinate system. In this way, the image pixel position information can be mapped to the two-dimensional coordinate position information. For example, Figure 3 The gray area in the figure uses the line extraction operator to extract the boundary line information, and the lower boundary of the extracted boundary line information is projected into the Cartesian coordinate system. Figure 4a shown.

[0118] See also Figure 4b As shown, Figure 4b This is a schematic diagram of calculating the boundary gradient. Using the sliding window size delta, the boundary gradient of the carpet boundary extraction line is calculated, which can be expressed mathematically as follows:

[0119] Dz=Δy / Δx,

[0120] Among them, Dz represents the boundary gradient, Δy represents the coordinate change of the y-axis in the sliding window delta, and Δx represents the coordinate change of the x-axis in the sliding window.

[0121] Obtain boundary gradients Dz = (Dz_1, Dz_2, ..., Dz_n) of n sliding windows of size Delta = (delta_1, delta_2, ..., delta_n); where n is a natural number greater than 1. The sizes of the sliding windows can be the same or different, and the positions of the sliding windows can be different. To improve detection accuracy, the sliding windows are continuously adjacent along the boundary. In this way, the boundaries of the carpet can be detected.

[0122] For any Dz_i in the boundary gradient Dz, if the Dz_i is greater than the set boundary gradient threshold sigma, it is considered that there is a risk in the boundary within the sliding window, and the risk boundary set R1 is obtained.

[0123] Eliminating the risk boundary, the low-risk carpet boundary is R2 = R–R1.

[0124] Since the boundary gradient is related to the coordinate change, the extracted boundary line can be projected onto the x-axis, see Figure 4c As shown in the figure, the thick black line segment is the projected boundary, most of which coincides with the x-axis. The grid part represents the carpet. In this way, for the flat boundary line part, the coordinate change of the y-axis in the sliding window is 0, and the boundary gradient is 0. For the uneven boundary line part, the coordinate change of the y-axis is obvious, and the boundary gradient is obvious. This can greatly reduce the amount of calculation and improve the accuracy of detection.

[0125] Step 203 : adopt different carpet placement strategies, ie, upward movement strategies, for different carpet categories, and adopt corresponding upward movement strategies to move from the boundary position of the low-risk area to the upper surface of the carpet.

[0126] As an example, according to the low-risk carpet boundary information, a carpet boundary position is selected as the upper carpet position information. For example, positions within a set range of the carpet boundary that are all low-risk carpet boundaries are selected as the upper carpet position.

[0127] If the adhesion between the carpet and the ground is strong, for example, silica gel carpet, then when the sweeper moves forward towards the carpet, its universal wheels or side brushes will not push the carpet.

[0128] When the carpet height is less than the first height threshold h_1, the carpet is placed in a straight-up manner, such as Figure 5 As shown in a,

[0129] When the carpet height is greater than or equal to the first height threshold h_1, the carpet is placed in an oblique manner, such as Figure 5 As shown in b;

[0130] If the carpet has weak adhesion to the ground, such as a thin pile carpet, when the sweeper moves forward towards the carpet, its universal wheels or side brushes will push the carpet.

[0131] When the carpet height is less than the second height threshold g_1, the carpet is placed upside down, such as Figure 5 As shown in c;

[0132] When the carpet height is greater than the second height threshold g_1, the carpet is placed in an inverted oblique manner, such as Figure 5 As shown in d.

[0133] Figure 5The medium-thick black arrow indicates the direction of movement of the sweeper.

[0134] in,

[0135] Forward driving means: the moving direction of the sweeper is perpendicular to the front of the sweeper.

[0136] The straight-up method for sweeping the carpet refers to a method in which the sweeper moves in the same direction as the normal direction of the front face of the sweeper and is perpendicular to the edge of the carpet. That is, when the normal direction of the front face of the sweeper points to the edge of the carpet and is perpendicular to the edge of the carpet, the sweeper moves in the same direction as the normal direction of the front face of the sweeper, so that the driving wheels of the sweeper move simultaneously to the upper surface of the carpet.

[0137] Obliquely ascending the carpet means that the moving direction of the sweeper is consistent with the normal direction of the front face of the sweeper, and the moving direction forms an angle with the edge of the carpet. That is to say, when the normal direction of the front face of the sweeper forms an angle with the edge of the carpet, the moving direction of the sweeper is consistent with the normal direction of the front face of the sweeper, so that the driving wheels of the sweeper move to the upper surface of the carpet one by one in sequence, rather than moving to the upper surface of the carpet at the same time.

[0138] The reverse-upward-upward method of sweeping the carpet refers to a method in which the sweeper moves in the opposite direction to the normal direction of the front face of the sweeper and is perpendicular to the edge of the carpet. That is, when the normal direction of the front face of the sweeper is away from the edge of the carpet and perpendicular to the edge of the carpet, the sweeper moves in the direction perpendicular to the edge of the carpet and away from the normal direction of the front face, so that the driving wheels of the sweeper move to the upper surface of the carpet at the same time.

[0139] Laying the carpet in an inverted oblique manner means that the moving direction of the sweeper is opposite to the normal direction of the front face of the sweeper, and the moving direction forms an angle with the boundary of the carpet. That is to say, when the normal direction of the front face of the sweeper forms an angle with the boundary of the carpet, the moving direction of the sweeper deviates from the normal direction of the front face of the sweeper, so that the driving wheels of the sweeper move to the upper surface of the carpet one by one in sequence, instead of moving to the upper surface of the carpet at the same time. In this way, the driving wheels move to the upper surface of the carpet in a rolling manner, which can avoid pushing the carpet.

[0140] The angle is between 5° and 85°, and preferably, the angle is 45°.

[0141] It should be understood that the carpet height can be the carpet thickness, which can be set according to the carpet type. For example, the height of the silica gel algae mud carpet is about 3 mm, the height of the thin pile carpet is less than 10 mm, and the height of the thick pile carpet is above 10 mm and below 20 mm.

[0142] As an example, after determining the upper carpet position information and the upward movement strategy, the sweeping robot determines the spatial position corresponding to the upper carpet position information based on the upper carpet position information, based on the current positioning results and map data, and moves to the upper surface of the carpet using the upward movement strategy at this position.

[0143] Step 204: After the sweeper moves to the upper surface of the carpet, it plans the cleaning path based on the acquired carpet area location information.

[0144] As an example, based on the acquired carpet area location information, the size information of the carpet surface area is determined, and a bow-shaped planning path is determined with the long side. Cleaning is performed according to the determined planning path to improve carpet cleaning efficiency.

[0145] The sweeper in this embodiment can stably move up to the top of the carpet and clean it efficiently, effectively avoiding the problem of the sweeper being unable to reach the carpet due to the sweeping brush, chassis, drive wheel and other structures. The carpet in this embodiment is universal and can be competent for cleaning carpets with different sweeper structural designs.

[0146] See also Figure 6 As shown, Figure 6 This is a schematic diagram of a surface cleaning device based on a mobile cleaning robot according to an embodiment of the present application. The device includes:

[0147] The surface covering information acquisition module is used to obtain the boundary information and surface covering category of the surface covering to be cleaned.

[0148] The low-risk boundary detection module is used to perform low-risk boundary detection on the boundary of the surface covering to be cleaned to determine low-risk boundary information. The low-risk boundary information is used to represent boundary information that is unlikely to cause the mobile cleaning robot to be unable to move to the surface covering.

[0149] The upward movement strategy determination module is used to determine the upward movement strategy of the mobile cleaning robot according to the acquired surface covering category, wherein the upward movement strategy ensures that the surface covering itself is not pushed during the process of the mobile cleaning robot moving the surface covering.

[0150] A movement control module is used to select a boundary position based on the low-risk boundary information, and move to the upper surface of the surface covering at the selected boundary position according to the determined upward movement strategy.

[0151] The task execution module is used to clean the surface of the surface covering according to the cleaning path planned by the surface covering.

[0152] As an example, the surface covering information acquisition module includes:

[0153] The boundary information acquisition module is used to use the trained deep learning segmentation network model to identify the current image data and obtain the surface covering area information in the current image data.

[0154] The surface covering category recognition module is used to use the trained deep neural network model to identify the surface covering area in the current image data and obtain the surface covering category.

[0155] See also Figure 7 As shown, Figure 7 Another schematic diagram of a surface covering cleaning device based on a mobile cleaning robot according to an embodiment of the present application. The device includes a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the computer program to implement any of the steps of the surface covering cleaning method based on a mobile cleaning robot.

[0156] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage. Alternatively, the memory may be at least one storage device located away from the processor.

[0157] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0158] As for the apparatus / network-side device / storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0159] In this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0160] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A surface covering cleaning method based on a mobile cleaning robot, characterized in that: The method includes: on the mobile cleaning robot side, Get the boundary information and surface covering type of the surface to be cleaned, Low-risk boundary detection is performed on the boundary of the surface covering to be cleaned to determine low-risk boundary information, where the low-risk boundary information is used to represent boundary information that is conducive to the mobile cleaning robot moving to the upper surface of the surface covering. Determine an upward movement strategy of the mobile robot according to the acquired surface covering category, wherein the upward movement strategy ensures that the surface covering itself is not pushed during the upward movement of the mobile cleaning robot to the upper surface of the surface covering. At the low-risk boundary position determined based on the low-risk boundary information, according to the determined upward movement strategy, move to the upper surface of the surface to be cleaned. On the surface of the surface to be cleaned, clean according to the cleaning path planned on the surface.

2. The surface covering cleaning method according to claim 1, wherein: The step of determining the upward movement strategy of the mobile robot according to the acquired surface covering category includes: Determine the strength of adhesion between the surface covering and the supporting surface and the height of the surface covering according to the type of surface covering. According to the strength of the adhesion between the surface covering and the supporting surface and the height of the surface covering, the corresponding upward movement strategy is determined. The upward movement strategy is used to determine the contact state between the walking wheels of the mobile cleaning robot and the boundary of the surface covering during the upward movement, and / or to determine the angle relationship between the moving direction of the mobile cleaning robot and the boundary of the surface covering during the upward movement.

3. The surface covering cleaning method according to claim 2, wherein: The mobile cleaning robot has a driving wheel and a driven wheel, wherein the driven wheel is located close to the front of the mobile cleaning robot, and the driving wheel is located away from the front of the mobile cleaning robot. The determining of the corresponding upward movement strategy according to the strength of the adhesion between the surface covering and the bearing surface and the height of the surface covering includes: In the case of strong adhesion between the surface covering and the bearing surface, If the height of the surface covering is not greater than the set first height threshold, the first method is adopted. If the height of the surface covering is greater than the set first height threshold, the second method is used. In cases where the adhesion between the surface covering and the bearing surface is weak, If the height of the surface covering is not greater than the set second height threshold, the third method is adopted. If the height of the surface covering is greater than the set second height threshold, the fourth method is adopted. in, The first method is: during the upward movement, each driving wheel of the mobile cleaning robot simultaneously contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering, and each driven wheel moves upward to the upper surface of the surface covering before the driving wheel contacts the boundary of the surface covering. The second method is: during the upward movement, each driving wheel of the mobile cleaning robot sequentially contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering, and each driven wheel contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering before the driving wheel. The third method is: during the upward movement, each driving wheel of the mobile cleaning robot simultaneously contacts the boundary of the surface covering and moves upward to the upper surface of the surface covering, and each driven wheel moves upward to the upper surface of the surface covering after the driving wheel contacts the boundary of the surface covering. The fourth method is: during the upward movement, each driving wheel of the mobile cleaning robot contacts the boundary of the surface covering one by one and moves up to the upper surface of the surface covering, and each driven wheel contacts the boundary of the surface covering and moves up to the upper surface of the surface covering later.

4. The surface covering cleaning method according to claim 3, wherein: The first method is a straight-up method, the second method is an oblique-up method, the third method is an inverted straight-up method, and the fourth method is an inverted oblique-up method. in, The above method is: the moving direction of the mobile cleaning robot is consistent with the normal direction of the front of the mobile cleaning robot, and the moving direction is perpendicular to the boundary of the surface covering. The oblique upward method is: the moving direction of the mobile cleaning robot is consistent with the normal direction of the front of the mobile cleaning robot, and the moving direction forms an angle with the boundary of the surface covering, The reverse and correct method is: the moving direction of the mobile cleaning robot is opposite to the normal direction of the front of the mobile cleaning robot, and the moving direction is perpendicular to the boundary of the surface covering. The inverted oblique upward method is that the moving direction of the mobile cleaning robot is opposite to the normal direction of the front of the mobile cleaning robot, and the moving direction forms an angle with the boundary of the surface covering.

5. The surface covering cleaning method according to claim 2, wherein: The mobile cleaning robot has only driving wheels but no driven wheels. The determining of the corresponding upward movement strategy according to the strength of the adhesion between the surface covering and the bearing surface and the height of the surface covering includes: In the case of strong adhesion between the surface covering and the bearing surface, If the height of the surface covering is not greater than the set first height threshold, the upward method is adopted. If the height of the surface covering is greater than the set first height threshold, the inclined upward method is adopted. In cases where the adhesion between the surface covering and the bearing surface is weak, If the height of the surface covering is not greater than the set second height threshold, the reverse-up method is adopted. If the height of the surface covering is greater than the set second height threshold, the reverse slant upward method is adopted. in, The above method is: the moving direction of the mobile cleaning robot is consistent with the normal direction of the front of the mobile cleaning robot, and the moving direction is perpendicular to the boundary of the surface covering. The oblique upward method is: the moving direction of the mobile cleaning robot is consistent with the normal direction of the front of the mobile cleaning robot, and the moving direction forms an angle with the boundary of the surface covering, The reverse and correct method is: the moving direction of the mobile cleaning robot is opposite to the normal direction of the front of the mobile cleaning robot, and the moving direction is perpendicular to the boundary of the surface covering. The inverted oblique upward method is that the moving direction of the mobile cleaning robot is opposite to the normal direction of the front of the mobile cleaning robot, and the moving direction forms an angle with the boundary of the surface covering.

6. The surface covering cleaning method according to claim 4 or 5, characterized in that: The angle is between 5° and 85°, The step of obtaining boundary information and surface covering category of the surface covering to be cleaned includes: The mobile cleaning robot obtains current image data while performing a cleaning task, and obtains regional information of the surface covering to be cleaned and the type of the surface covering from the current image data; The cleaning of the upper surface of the surface covering to be cleaned according to the cleaning path planned for the surface covering includes: Surface covering map data is acquired from the map data, and based on the surface covering map data, a bow-shaped cleaning path is planned with the long side of the surface covering so as to perform cleaning according to the cleaning path.

7. The surface covering cleaning method according to claim 6, wherein: The angle is 45°, The step of obtaining the area information of the surface covering to be cleaned and the type of the surface covering from the current image data includes: The trained deep learning segmentation network model is used to identify the current image data and obtain the surface covering area information in the current image data. Using the trained deep neural network model, the surface covering area in the current image data is identified to obtain the surface covering category; The determining of low-risk boundary information based on the acquired surface covering area information includes: Extracting the boundary information of the surface covering from the acquired surface covering area information, For the extracted surface covering boundary, the boundary gradient in more than one sliding window is calculated to obtain more than one boundary gradient. For each boundary gradient, if the boundary gradient is not greater than a set boundary gradient threshold, the boundary in the sliding window corresponding to the boundary gradient is determined to be a low-risk boundary.

8. The surface covering cleaning method according to claim 7, wherein: There are multiple sliding windows, each of which has the same size, and each of which is continuously adjacent to the other along the boundary. The boundary gradient is calculated as follows: The ratio between the one-dimensional position change and the other-dimensional position change of the extracted surface covering boundary within a sliding window is calculated to obtain the boundary gradient within the sliding window.

9. A surface covering cleaning device based on a mobile cleaning robot, characterized in that: The device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the computer program to implement the steps of the surface covering cleaning method based on a mobile cleaning robot according to any one of claims 1 to 8.

10. A mobile cleaning robot, characterized in that: It comprises a surface covering cleaning device based on a mobile cleaning robot as described in claim 9.

Images