Cleaning robot path planning method and system based on dynamic obstacle prediction

Abstract

The invention belongs to the technical field of cleaning robots, and particularly relates to a cleaning robot path planning method and system based on dynamic obstacle prediction. The method comprises the following steps: acquiring first image information, calculating an obstacle area proportion to determine a region boundary line and divide a cleaning region, and then constructing a path grid to generate a cleaning path; in the process of executing the cleaning task, the time change rate of the real-time relative distance between the movable obstacle and the cleaning path is calculated, and a cleaning path offset instruction or a pause instruction is output; and after the cleaning task is executed, comparing the first image information with the second image information to calculate a correction offset, and correcting the cleaning path according to the correction offset. The sweeping coverage comprehensiveness of a complex area is improved through self-adaptive area division, the operation safety and continuity are improved through predictive dynamic obstacle avoidance, and the long-term accuracy and sweeping quality of a path are ensured through a double closed-loop correction mechanism.

Description

Technical Field

[0001] This invention belongs to the field of cleaning robot technology, specifically relating to a cleaning robot path planning method and system based on dynamic obstacle prediction. Background Technology

[0002] Autonomous mobile robots, represented by cleaning robots, can efficiently and autonomously complete cleaning tasks in designated areas in scenarios such as commercial cleaning and industrial automation through autonomous navigation and path planning, thereby freeing up manpower. The quality of a cleaning robot's environmental perception and intelligent decision-making capabilities directly determines its level of automation, operational safety, and user experience.

[0003] Current robotic vacuum cleaners mostly rely on preset maps or fixed paths for operation, resulting in poor environmental adaptability and difficulty in responding to dynamic elements in the environment, such as temporarily placed items, moving pedestrians, or pets. This leads to rigid path planning, ineffective cleaning, task interruption, or insufficient coverage. Current obstacle perception technologies have significant limitations. Infrared sensor-based solutions are prone to missing detection of transparent or highly reflective objects such as glass and mirrors, leading to collision damage. Ultrasonic-based solutions have low spatial resolution and struggle to identify the outlines and geometric shapes of obstacles, especially when dealing with dense, small obstacle combinations. These obstacles can easily be misjudged as an impassable area, resulting in overly conservative path planning and low cleaning efficiency. Current robotic vacuum cleaners can only avoid obstacles as simple geometric entities, unable to identify dynamic objects or predict their trajectories. This means that the decision-making logic of current robotic vacuum cleaners remains at the passive reaction level, which can easily lead to safety issues.

[0004] To address the aforementioned problems, this invention provides a path planning method and system for cleaning robots based on dynamic obstacle prediction. Summary of the Invention

[0005] The purpose of this invention is to provide a path planning method and system for cleaning robots based on dynamic obstacle prediction, so as to improve the cleaning efficiency and path planning rationality of cleaning robots, and to prevent the phenomenon of untimely offset correction due to the movement of obstacles.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A path planning method for a cleaning robot based on dynamic obstacle prediction includes:

[0008] Based on the first image information of the cleaning area, a cleaning path is generated that covers the non-obstacle areas within the cleaning area;

[0009] During the cleaning task along the cleaning path, when the time change rate of the real-time relative distance between the movable obstacle in the cleaning area and the cleaning path meets the preset path control trigger condition, a cleaning path offset command or a pause command is output based on the time change rate.

[0010] After the cleaning task is completed, the second image information of the cleaning area is obtained, and the correction offset is calculated based on the image difference between the second image information and the first image information to correct the cleaning path.

[0011] The path control triggering conditions include: when the time change rate is greater than the first preset threshold, an output cleaning path offset command is triggered; when the time change rate is less than the second preset threshold, an output pause command is triggered.

[0012] Preferably, the step of generating a cleaning path covering non-obstacle areas within the cleaning area based on the first image information of the cleaning area includes:

[0013] The area boundary line is determined based on the area ratio of obstacles in the first image information; the cleaning area is divided based on the area boundary line; the boundary contour line of the cleaning area defined by the area boundary line is obtained based on the distribution state of obstacles in the first image information, and the cleaning start point is set on the boundary contour line to generate the cleaning path.

[0014] Preferably, the step of generating a cleaning path covering non-obstacle areas within the cleaning area based on the first image information of the cleaning area further includes:

[0015] Starting from the cleaning start point, a path grid covering the non-obstacle areas within the cleaning area is constructed, and the intersections of the path grid are determined as cleaning path points. The cleaning path is generated by connecting the cleaning path points according to the preset cleaning sequence.

[0016] Preferably, after outputting the pause command, the method further includes:

[0017] When the movement state of a movable obstacle remains unchanged within a preset time period, the movable obstacle is re-identified as an immovable obstacle; the current image information centered on the stopping point of the cleaning robot is collected and compared with the first image information to calculate the image difference degree; when the image difference degree is less than the preset tolerance value, the instruction to continue executing the original cleaning path is output, or when the image difference degree is greater than or equal to the preset tolerance value, the cleaning path is replanned.

[0018] Preferably, the step of calculating the correction offset based on the image difference between the second image information and the first image information includes:

[0019] The first and second image information are normalized, and the positions of the first and second sets of feature points are extracted respectively; the positional deviation between the first and second sets of feature points is calculated; and the maximum value among all positional deviation values ​​is determined as the correction offset.

[0020] Preferably, the method further includes:

[0021] During the cleaning process, the movement of obstacles in the cleaning area is monitored in real time. When the position of an obstacle changes, it is identified as a movable obstacle, and the position change of the movable obstacle is recorded to obtain its movement trajectory, thereby calculating the real-time relative distance.

[0022] 7. A path planning system for a cleaning robot based on dynamic obstacle prediction, comprising:

[0023] The environmental recognition module is used to acquire the first image information of the cleaning area and identify obstacles within it;

[0024] The initial path planning module is used to generate a cleaning path that covers non-obstacle areas within the cleaning area based on the recognition results of the environment recognition module.

[0025] The dynamic obstacle monitoring module is used to monitor the movement status of movable obstacles in the cleaning area in real time during the cleaning task along the cleaning path;

[0026] The dynamic path decision module is configured to respond to the monitoring results of the dynamic obstacle monitoring module. It is used to calculate the time change rate of the real-time relative distance between the movable obstacle and the cleaning path based on the movement state of the movable obstacle. When the time change rate meets the preset path control trigger condition, it outputs a cleaning path offset command or a pause command according to the time change rate.

[0027] The path correction module is used to calculate the correction offset to correct the cleaning path by comparing the first image information with the second image information obtained after cleaning after the cleaning task is completed.

[0028] Preferably, generating a cleaning path that covers non-obstacle areas within the cleaning area includes:

[0029] The area boundary line is determined based on the area ratio of obstacles in the first image information; the cleaning area is divided based on the area boundary line; the boundary contour line of the cleaning area defined by the area boundary line is obtained based on the distribution state of obstacles in the first image information, and the cleaning start point is set on the boundary contour line to generate the cleaning path.

[0030] Starting from the cleaning start point, a path grid covering the non-obstacle areas within the cleaning area is constructed, and the intersections of the path grid are determined as cleaning path points. The cleaning path is generated by connecting the cleaning path points according to the preset cleaning sequence.

[0031] Preferably, the path control triggering conditions include: when the time change rate is greater than a first preset threshold, triggering an output cleaning path offset command; when the time change rate is less than a second preset threshold, triggering an output pause command.

[0032] Preferably, the step of calculating the correction offset to correct the cleaning path by comparing the first image information with the second image information obtained after cleaning includes: normalizing the first image information and the second image information, and extracting the positions of the first group of feature points and the second group of feature points respectively; calculating the positional deviation value between the first group of feature points and the second group of feature points; and determining the maximum value among all positional deviation values ​​as the correction offset.

[0033] Beneficial effects

[0034] This invention acquires first image information, calculates the area ratio of obstacles, and determines the boundary lines of the area to divide the cleaning area. In the divided non-obstacle areas, a path grid is constructed to generate a cleaning path. By performing more refined path planning for complex areas with dense obstacles, this invention overcomes the limitations of traditional fixed planning methods and improves the rationality of the initial cleaning path and the comprehensiveness of cleaning coverage of complex areas.

[0035] When performing a cleaning task, this invention identifies movable obstacles by monitoring their movement in real time. It calculates the rate of change of the real-time relative distance between the obstacle and the cleaning path and predicts the obstacle's future trajectory to assess the collision risk and output a cleaning path deviation command or a pause command. For moving obstacles of different threat levels, it adopts a predictive avoidance or temporary pause strategy to avoid collisions, ensure operational safety, reduce unnecessary path interruptions caused by low-risk interference, and improve the continuity and execution efficiency of the cleaning task in a dynamic environment.

[0036] After the task is completed, the present invention calculates the correction offset by comparing the first and second image information to correct the cleaning path and adapt to static changes in the environment. After the robot pauses due to a movable obstacle, it compares the current image information of the robot's pause point with the first image information to determine whether the path needs to be replanned in order to cope with temporary changes in the task. This dynamically maintains the accuracy of the path, overcomes the problem of path failure caused by static or temporary changes in the environment, and improves the long-term coverage integrity and cleaning quality of the target area. Attached Figure Description

[0037] Figure 1 This is a flowchart of the method provided by the present invention;

[0038] Figure 2 This is a flowchart of the present invention, which outputs a cleaning path offset instruction or a pause instruction based on the time change rate.

[0039] Figure 3 This is a flowchart of the present invention for correcting the cleaning path based on the correction offset;

[0040] Figure 4 This is a block diagram of the system provided by the present invention. Detailed Implementation

[0041] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0042] Example 1

[0043] Please refer to Figures 1-3 This embodiment provides a path planning method for a cleaning robot based on dynamic obstacle prediction, which specifically includes the following steps:

[0044] Image acquisition and region segmentation are performed. Digital image data representing the initial environmental state of the cleaning area, acquired by the image acquisition device on the cleaning robot, is used as the first image information of the cleaning area. The first image information is input into the cleaning area recognition and processing flow. This recognition and processing flow analyzes the first image information to identify obstacles. Through image processing steps, such as edge detection, color and texture analysis, the set of pixels in the image that have significant differences in color, texture or height from the ground background is identified as pixel regions that meet the visual characteristics of obstacles. For each identified obstacle region, an image feature vector is constructed. The image feature vector is a set of structured numerical data used to describe the geometric and positional attributes of the obstacle, including the coordinate point set of the obstacle outline, the size of the circumscribed rectangle, the centroid coordinates, and the total number of pixels. Based on the image feature vector, the area ratio of obstacles in the cleaning area is calculated.

[0045] Specifically, the steps for calculating the area ratio of obstacles in the cleaned area are as follows: based on the total number of obstacle pixels recorded in the image feature vector, count the number of pixel units occupied by all obstacles in the first image information, and compare this number with the total number of pixel units in the first image information. This ratio is then determined as the area ratio of obstacles, providing a quantitative basis for subsequent environmental complexity assessment.

[0046] The functional relationship for calculating the area ratio of obstacles is:

[0047]

[0048] In the formula, The area ratio of obstacles represents the proportion of the image area occupied by obstacles to the total image area. It is a numerical indicator for assessing the complexity of the environment. This represents the total number of pixels representing obstacles, which is the sum of the number of all pixel units identified as obstacles in the first image information. Represents the total number of pixels, which means the total number of pixel units in the first image information, i.e., the product of the width and height of the image.

[0049] The boundary lines used to delineate the cleanable area are determined based on the area ratio of obstacles. These boundary lines are one or more closed or open line segments generated according to the location and distribution of obstacles, used to accurately define the area that the cleaning robot can pass through. Specifically, determining the boundary lines for delineating the cleanable area includes: matching a sampling point density value to the area ratio of obstacles according to a preset correspondence. The sampling point density value determines the accuracy of the boundary line's fit to the shape of the obstacle. If the correspondence is monotonically increasing, that is, areas with a higher area ratio of obstacles correspond to a higher sampling point density, a series of closely fitting coordinate points are generated at the edge of the obstacle using this sampling point density, and these coordinate points are connected to form an accurate boundary line. Conversely, in areas with sparse obstacles, a lower sampling point density is used to improve processing efficiency. Furthermore, if the calculation finds that the area of ​​a densely obstructed area is greater than a preset area threshold, multiple boundary lines are set inside or around that area, thereby dividing the original cleaning area into several independent sub-areas, and performing independent obstacle identification and path planning for each sub-area.

[0050] Furthermore, path planning is performed to generate a cleaning path covering the non-obstacle areas within the cleaning area. Based on the obstacle distribution identified in the first image information, the closed polygon defined by the generated area boundary line, used to define the outermost perimeter of the entire cleanable area, is obtained as the boundary contour line of the cleaning area. A specific geographic coordinate of a point on this boundary contour line is selected as the cleaning starting point. Starting from the cleaning starting point, a path grid that can cover all non-obstacle areas within the cleaning area is constructed according to predetermined geometric generation rules. Within the coordinate system of the cleaning area, a set of equally spaced horizontal parallel lines and a set of vertical parallel lines are generated to form a path grid. The initial grid lines are then removed or truncated, eliminating all grid segments that intersect with the outlines of identified immovable obstacles. The remaining valid segments form the path grid, a network structure of intersecting horizontal and vertical valid segments that covers all accessible cleaning areas. The geometric intersections of all horizontal and vertical valid segments in the path grid are obtained as path intersections, and the set of these path intersections is determined as a set of cleaning path points. Each point in the set of path intersections is a target location that the cleaning robot needs to reach. This set of path points collectively covers all accessible non-obstacle areas, ensuring the integrity of the cleaning path.

[0051] Furthermore, the cleaning task is executed by transforming the discrete set of path points into a continuous and efficient single cleaning path according to the preset connection logic. In this embodiment, the connection logic is a bow-shaped traversal or the nearest neighbor principle, which connects the generated cleaning path points in an orderly manner, thereby generating a complete cleaning path that defines the complete travel route of the cleaning robot, consisting of a cleaning start point, a series of ordered cleaning path points, and an end point. This path is then output to the motion control unit of the cleaning robot. The cleaning robot, according to the cleaning sequence set by the cleaning path, starts from the cleaning start point and proceeds to each cleaning path point in sequence until it reaches the end point. The end point marks the end of the predetermined cleaning task. During the cleaning task along the cleaning path, the position information of the cleaning robot is recorded in real time to form an actual trajectory. In this embodiment, the actual trajectory is a time series of the cleaning robot's position coordinates recorded in real time by an odometer or GPS, and the actual arrival time of each cleaning path point is recorded for subsequent analysis and optimization.

[0052] Furthermore, during the cleaning process, dynamic obstacle detection and prediction are performed. By continuously comparing image information collected at different time points, the position status of obstacles in the cleaning area is monitored in real time. If the position coordinates of an obstacle are detected to change continuously and regularly, the obstacle is determined to be a movable obstacle. The orderly set of position coordinates at a series of consecutive time points is recorded to form a movement trajectory. After obtaining the movement trajectory of the movable obstacle, a predictive calculation process is performed. This process estimates the current movement speed and direction of the obstacle based on a series of recent historical position data. Based on this, the most likely path of the movable obstacle, which is formed by connecting the predicted positions in a preset time period in the future, is calculated as the predicted trajectory. Then, it is determined whether the predicted trajectory intersects with the cleaning path currently planned by the cleaning robot.

[0053] The functional relationship of the predictive computation process is:

[0054]

[0055] In the formula, This represents the predicted position vector, which indicates the location of the obstacle in the future. Predicted spatial coordinates after time; This represents the current position vector, indicating the position of the obstacle at the current time point. Spatial coordinates; This represents the previous position vector, indicating the position of the obstacle at the previous time point. Spatial coordinates; It represents the time increment, which means the time span for predicting the future; This represents the current time point, meaning the current location vector has been collected. timestamp; This indicates the previous time point, meaning the location vector was collected at the previous time. Timestamp.

[0056] Specifically, if it is determined that the predicted trajectory intersects with the cleaning path currently planned by the cleaning robot, the current movement direction of the movable obstacle is determined as the critical direction, and a temporary key monitoring area is constructed with the predicted intersection as the center. Within this area, the image acquisition frequency can be increased to perform more precise tracking and achieve precise monitoring of potential threats. If it is determined that there is no intersection, the current movement direction of the movable obstacle is determined as the non-critical direction, and the cleaning robot can maintain the current cleaning path and maintain the normal monitoring state. When the position of the movable obstacle does not change within a preset time period or its position change is less than a preset small displacement threshold, in this embodiment, the total length of the movable obstacle's movement path within 1 minute is less than 10mm, its state is updated, it is re-identified as an immovable obstacle, and its final stationary position is recorded.

[0057] Furthermore, during or after the cleaning task, dynamic path optimization is performed. During the cleaning process, the actual trajectory of the cleaning robot is continuously compared with the cleaning path. If the distance between the two at a certain position exceeds a preset tolerance value, the actual trajectory segment exceeding the preset tolerance value is defined as a path offset segment, and its position and range are recorded. After the cleaning task is completed, second image information of the cleaning area is acquired. This second image information reflects environmental changes or evaluates the cleaning effect. To ensure the consistency of the comparison benchmark, the first and second image information are normalized. Normalization adjusts their brightness and contrast to a uniform standard range. A set of highly discriminative and stable first set of feature points is extracted from the normalized first image information. This first set of feature points is the first image... The pixels extracted from the information that are significant, stable, and reproducible can be corner points of fixed objects, unique texture regions, or scale-invariant feature transformation key points. In the normalized second image information, a feature matching algorithm is used to find a second set of feature points that are in the same physical location as the first set of feature points. The Euclidean distance or vector difference between each pair of corresponding first and second set feature points is calculated as the coordinate position deviation value to obtain a set of deviation data. The maximum value among all coordinate position deviation values ​​is determined as the correction offset. This correction offset represents the maximum displacement between two images and can be used to correct the original cleaning path, for example, by overall translation or local adjustment of path points, to plan the supplementary cleaning task and cover cleaning areas that may be missed due to path offset.

[0058] Furthermore, dynamic obstacle processing and command output are performed. The movement trajectory of the movable obstacle is defined as a detection curve. Based on the detection curve and the cleaning path, the real-time relative distance between the movable obstacle and the cleaning robot is calculated in real time. The real-time relative distance over time is further calculated, and the rate of change over time is compared with a preset threshold to output corresponding control commands.

[0059] The functional relationship for calculating the rate of change over time is:

[0060]

[0061] In the formula, The rate of change over time represents the percentage change at a given time point. The instantaneous rate of change of the distance between the robot and the obstacle; a negative value indicates that they are getting closer and a positive value indicates that they are moving away. This indicates the current relative distance, meaning at a given time point. At that time, the Euclidean distance between the robot and the obstacle; This indicates the previous relative distance, meaning at the previous point in time. At that time, the Euclidean distance between the robot and the obstacle; This indicates the current point in time, signifying the measurement of the current relative distance. timestamp; This indicates the previous point in time, meaning the relative distance was measured at the previous point in time. Timestamp.

[0062] Specifically, when the absolute value of the time change rate is greater than a first preset threshold representing that an obstacle is rapidly approaching and needs to be avoided immediately, a collision risk is determined, and a cleaning path offset command is immediately output to make the cleaning robot change its route to avoid the obstacle. When the absolute value of the time change rate is less than a second preset threshold representing that the distance between the obstacle and the robot is changing slowly and can be paused, a pause command is output, and the cleaning robot stops at its current position, which is recorded as the pause point. After the pause command is output, the movable obstacle is continuously monitored. When its movement state remains unchanged within a preset time period, it is re-identified as an immovable obstacle, and the position of the pause point is defined as a verification point. The verification point serves as the reference point for a new round of image acquisition and comparison. The current image information centered on the verification point is acquired and compared with the corresponding area in the first image information acquired at the start of the task. The image difference is calculated through comparison rules. This calculation may include a quantitative comparison of the brightness, contrast, and structural information of the corresponding areas of the two images, and finally a single value is obtained.

[0063] The functional relationship for calculating image dissimilarity is as follows:

[0064]

[0065] In the formula, This indicates the image difference, which quantifies the overall difference in brightness and contrast between two image regions. The larger the value, the more significant the difference. The original image region refers to the corresponding area centered on the verification point in the first image information. This indicates the current image region, meaning the image region acquired centered on the verification point. This represents the average pixel intensity of the original region, which is the meaning of the image region. The arithmetic mean of the grayscale values ​​of all pixels within the area reflects the overall brightness of the region; This represents the average pixel intensity of the current region, which is the meaning of the image region. The arithmetic mean of the grayscale values ​​of all pixels within the area reflects the overall brightness of the region; This represents the standard deviation of pixel intensity in the original region, which indicates the image region. The standard deviation of the grayscale values ​​of all pixels within a region reflects the contrast or texture complexity of the area. This represents the standard deviation of pixel intensity in the current region, which indicates the image region. The standard deviation of the grayscale values ​​of all pixels within a region reflects the contrast or texture complexity of the area. This represents the weighting coefficient, which is a preset constant used to adjust the importance of brightness and contrast differences in the final difference calculation.

[0066] Specifically, if the image difference is less than the preset tolerance value, it indicates that the environment has not changed significantly, and the instruction to continue executing the original cleaning path is output. If the image difference is greater than or equal to the preset tolerance value, it indicates that the position of the obstacle has been fixed and has affected the original path. At this time, based on the updated obstacle position information and combined with the correction offset, the path planning process will be called to regenerate a segment or the entire cleaning path.

[0067] Example 2

[0068] Please refer to Figure 4 This embodiment provides a path planning system for a cleaning robot based on dynamic obstacle prediction, which can avoid movable obstacles in real time and adaptively correct the path after the cleaning task is completed, thereby improving the cleaning efficiency and safety of the cleaning robot in complex and dynamic environments.

[0069] In its specific implementation, this system can be integrated into the main control unit of a cleaning robot. The main control unit includes, but is not limited to, at least one processor and a memory communicatively connected to at least one processor. The system specifically includes the following modules:

[0070] The environmental recognition module is configured to acquire a global image of the cleaning area using visual sensors such as a wide-angle camera mounted on the cleaning robot before the cleaning task begins. This image is then used as the first image information and analyzed to identify non-obstacle areas such as floors and carpets, as well as obstacles such as walls and fixed objects. This identification result provides a basic environmental map for subsequent initial path planning.

[0071] The initial path planning module receives obstacle information provided by the environment recognition module and generates a cleaning path covering the non-obstacle areas within the cleaning area. Based on the area ratio of obstacles in the first image information, it determines the boundary lines of the region used for macroscopic segmentation, thereby dividing the entire cleaning area into several sub-regions for cleaning one by one. It further analyzes the specific distribution of obstacles in the first image information to obtain the boundary contour lines of the non-obstacle areas that can be cleaned, which are precisely defined by the region boundary lines. On this boundary contour line, a cleaning start point is set. Starting from the cleaning start point, a virtual path grid is constructed in the non-obstacle area, and the path intersections of the path grid are determined as a series of cleaning path points. According to the preset cleaning sequence such as bow shape or spiral shape, the cleaning path points are connected in sequence to generate a complete and executable initial cleaning path.

[0072] The dynamic obstacle monitoring module is activated during the cleaning robot's cleaning task along the cleaning path. It uses a vision sensor to monitor environmental changes in the cleaning area in real time. When the position of an obstacle in the area changes compared to its position in the first image information, the obstacle is identified as a movable obstacle. The module continuously records the position change of the movable obstacle to obtain its movement trajectory. Based on the obtained movement trajectory and the cleaning robot's own cleaning path, the module calculates the real-time relative distance between the cleaning robot and the movable obstacle and transmits this information to the dynamic path decision module.

[0073] The dynamic path decision module operates in response to the monitoring results of the dynamic obstacle monitoring module. It receives the real-time relative distance and calculates the rate of change over time. This rate of change over time reflects the approaching or moving-away speed between the cleaning robot and the movable obstacle. It makes decisions based on preset path control trigger conditions. When the absolute value of the rate of change over time is greater than a first preset threshold, it indicates a high risk of collision. At this time, a cleaning path offset command is triggered to guide the cleaning robot to adjust its trajectory in advance to detour. When the absolute value of the rate of change over time is less than a second preset threshold, it may indicate that the obstacle has stopped or is moving slowly and blocking the path. At this time, a pause command is output to make the cleaning robot pause and wait.

[0074] After the pause command is output, the dynamic path decision module performs subsequent processing. If the movement status of a movable obstacle remains unchanged within a preset time period, it is temporarily re-identified as an immovable obstacle. The current image information centered on the pause point is collected and compared with the first image information to calculate the image difference. If the image difference is less than the preset tolerance value, it indicates that the environment has not changed much, and the command to continue executing the original cleaning path will be output. Conversely, if the image difference is greater than or equal to the preset tolerance value, it indicates that the environment has changed significantly, and the initial path planning module will be triggered to re-plan the cleaning path based on the current position and environment.

[0075] The path correction module is activated after the entire cleaning task is completed. It acquires the second image information of the cleaning area and calculates the correction offset by comparing the second image information with the first image information before the task started. This is used to correct the cleaning path generated in this case and provide a more accurate reference for the next task. The first and second image information are normalized to eliminate differences such as lighting and shooting angle. A set of stable and corresponding feature points are extracted from the two images, namely the first set of feature points and the second set of feature points. By calculating the coordinate position deviation value between the two sets of corresponding feature points, a series of deviation data can be obtained. The maximum value among all coordinate position deviation values ​​is determined as the correction offset, and this offset is used to globally correct the entire cleaning path map.

[0076] Through the coordinated work of the above modules, the robot can plan the initial cleaning path, predict and avoid moving obstacles, and perform self-learning and path correction after the task is completed, thereby improving the robot's adaptability and work efficiency in dynamic and complex environments.

[0077] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from it. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1.A method for path planning of a cleaning robot based on dynamic obstacle prediction, characterized in that, The method comprises the following steps: Based on the first image information of the cleaning area, a cleaning path covering the non-obstacle area in the cleaning area is generated; During the execution of the cleaning task along the cleaning path, when the time variation rate of the real-time relative distance between the movable obstacle in the cleaning area and the cleaning path meets the preset path control trigger condition, a cleaning path offset instruction or a pause instruction is output based on the time variation rate; After the cleaning task is completed, the second image information of the cleaning area is obtained, and the correction offset is calculated according to the image difference between the second image information and the first image information to correct the cleaning path; The path control trigger condition includes: when the time variation rate is greater than a first preset threshold, the cleaning path offset instruction is triggered to be output; when the time variation rate is less than a second preset threshold, the pause instruction is triggered to be output. 2.The method of claim 1, wherein, The step of generating the cleaning path covering the non-obstacle area in the cleaning area based on the first image information of the cleaning area comprises: Based on the obstacle area ratio in the first image information, the area boundary line is determined; the cleaning area is divided based on the area boundary line; the boundary contour line of the cleaning area defined by the area boundary line is obtained based on the distribution state of the obstacle in the first image information, and the cleaning starting point is set on the boundary contour line to generate the cleaning path. 3.The method of claim 2, wherein, The step of generating the cleaning path covering the non-obstacle area in the cleaning area based on the first image information of the cleaning area further comprises: Taking the cleaning starting point as the starting point, the path grid covering the non-obstacle area in the cleaning area is constructed, and the path intersection point of the path grid is determined as the cleaning path point, so that the cleaning path is generated in the manner of connecting the cleaning path points according to the preset cleaning sequence. 4.The method of claim 1, wherein, After outputting the pause instruction, the method further comprises: When the moving state of the movable obstacle remains unchanged within a preset time period, the movable obstacle is re-identified as an immovable obstacle; the current image information centered on the pause point of the cleaning robot is collected and compared with the first image information to calculate the image difference; when the image difference is less than a preset tolerance value, an instruction to continue executing the original cleaning path is output, or when the image difference is greater than or equal to the preset tolerance value, the cleaning path is re-planned. 5.The path planning method of a cleaning robot based on dynamic obstacle prediction according to claim 1, wherein, The step of calculating the correction offset according to the image difference between the second image information and the first image information comprises: The first image information and the second image information are normalized, and the positions of the first group of feature points and the second group of feature points are extracted respectively; the position deviation value between the first group of feature points and the second group of feature points is calculated; the maximum value of all position deviation values is determined as the correction offset. 6.The path planning method of a cleaning robot based on dynamic obstacle prediction according to claim 1, wherein, The method further comprises: During the execution of the cleaning task, the moving state of the obstacle in the cleaning area is monitored in real time, and when the position of the obstacle changes, the obstacle is determined as a movable obstacle, and the position change of the movable obstacle is recorded to obtain the moving trajectory, so as to calculate the real-time relative distance. 7.A path planning system for a cleaning robot based on dynamic obstacle prediction, characterized by, The method comprises the following steps: An environment recognition module is used to obtain the first image information of the cleaning area and recognize the obstacles therein; An initial path planning module is configured to generate a cleaning path covering the non-obstacle area in the cleaning area based on the recognition result of the environment recognition module; A dynamic obstacle monitoring module is configured to monitor the moving state of the movable obstacle in the cleaning area in real time during the execution of the cleaning task along the cleaning path; A dynamic path decision module is configured in response to the monitoring result of the dynamic obstacle monitoring module, and is configured to calculate the time variation rate of the real-time relative distance between the movable obstacle and the cleaning path based on the moving state of the movable obstacle, and output a cleaning path offset instruction or a pause instruction according to the time variation rate when the time variation rate meets a preset path control trigger condition; A path correction module is configured to calculate a correction offset to correct the cleaning path by comparing the first image information with the second image information obtained after cleaning after the execution of the cleaning task. 8.The path planning system of a cleaning robot based on dynamic obstacle prediction according to claim 7, wherein, The generation of the cleaning path covering the non-obstacle area in the cleaning area includes: determining the area boundary line based on the obstacle area ratio in the first image information, dividing the cleaning area based on the area boundary line, obtaining the boundary contour line of the cleaning area defined by the area boundary line based on the distribution state of the obstacle in the first image information, and setting the cleaning starting point on the boundary contour line to generate the cleaning path; starting from the cleaning starting point, constructing a path grid covering the non-obstacle area in the cleaning area, and determining the path intersection points of the path grid as cleaning path points to generate the cleaning path in the manner of connecting the cleaning path points according to the preset cleaning sequence. 9.The path planning system of a cleaning robot based on dynamic obstacle prediction of claim 7, wherein, The path control trigger condition includes: when the time variation rate is greater than a first preset threshold, the cleaning path offset instruction is triggered; and when the time variation rate is less than a second preset threshold, the pause instruction is triggered. 10.The path planning system of a cleaning robot based on dynamic obstacle prediction of claim 7, wherein, The calculation of the correction offset to correct the cleaning path by comparing the first image information with the second image information obtained after cleaning includes: normalizing the first image information and the second image information, and extracting the positions of the first group of feature points and the second group of feature points, respectively; calculating the position deviation value between the first group of feature points and the second group of feature points; and determining the maximum value of all position deviation values as the correction offset.

Images