Full-coverage cleaning path planning method based on Bezier curve optimization

Abstract

The invention discloses a full-coverage cleaning path planning method based on Bezier curve optimization, relates to the technical field of robot path planning, and solves the problems that in the prior art, path smoothness and full-coverage leakproofness are difficult to consider at the same time, a boundary is easy to leak, and scene adaptability is insufficient. The method comprises the following steps: the area preprocessing module performs rotation correction on a cleaning area; the path generation module adaptively selects a zigzag or rectangular-ambulatory-plane algorithm to generate a basic broken line path according to the regional features; the path optimization module dynamically selects a Bezier curve order to perform smoothing processing by extracting a key inflection point and a region center as control points; and the path adjustment module inversely transforms the optimized path to an original coordinate system, supplements a boundary path and joins the boundary path, and finally generates an executable cleaning path. According to the method, on the premise of ensuring tight coverage of the region, the path smoothness is remarkably improved, the edge scanning missing gap is effectively eliminated, and the adaptability to different scenes is enhanced through a self-adaptive strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot path planning, in particular to a full coverage path planning method for a cleaning robot, and more particularly to a path planning method combining Bezier curve optimization and region adaptive preprocessing to balance path smoothness, full coverage rate and scene adaptability. BACKGROUND

[0002] With the development of service robot technology, cleaning robots have been widely used in home, office, industrial workshop and other scenes. Path planning technology is one of the core functions of a cleaning robot, which directly determines the cleaning efficiency, coverage rate and equipment service life.

[0003] The current mainstream cleaning path planning technology mainly includes grid method, A algorithm and its improved algorithm, genetic algorithm, etc. The grid method quickly judges the feasible region by discretizing the environment into grids, but the generated path is mostly a broken line, which has poor smoothness and the robot frequently turns, which easily causes increased energy consumption and component wear. Although the A algorithm and its improved algorithm can quickly search for the optimal path, they are prone to path redundancy in full coverage scenarios and lack adaptability to complex polygon regions. Although intelligent optimization algorithms such as genetic algorithm can optimize path parameters, they have high computational complexity and poor real-time performance.

[0004] In order to improve the path smoothness, some existing technologies attempt to introduce curve fitting methods, such as using Bezier curves or B-spline curves to smooth the broken line path generated based on the grid method or A* algorithm. However, the applicant found that in the specific application scenario of full coverage cleaning, directly applying such general curve smoothing techniques would introduce new technical problems: since curve fitting aims to pursue global smoothness, it often ignores the strong constraint requirement of cleaning tasks on the rigorous coverage of region boundaries and corners. The path after curve smoothing may deviate from the original rigorous coverage trajectory, resulting in "gaps" that cannot be covered or "overflow" regions of the trajectory near the region edges, obstacles and complex polygon corners, thereby causing missed cleaning. In other words, the existing technology has not effectively solved the inherent contradiction between "path smoothness optimization" and "full coverage rigorousness constraint".

[0005] At the same time, the existing technology lacks a unified framework that can automatically adapt the path generation strategy according to the region shape and ensure that the final smoothed path does not destroy the coverage integrity when dealing with different shapes of cleaning regions (such as rectangles, polygons, and complex regions with obstacles), which limits the cleaning effect and lacks flexibility and adaptability. SUMMARY

[0006] In order to improve the technical problems in the existing cleaning robot path planning method, that is, the path smoothness and the region full coverage tightness cannot be considered together due to direct application of general curve smoothing technology, the clearance is easy to be missed in the corner of the boundary, and the self-adaptive ability to different scenes is lacked, a full coverage cleaning path planning method based on Bezier curve optimization is provided.

[0007] The application provides a full coverage cleaning path planning method based on Bezier curve optimization, which adopts the following technical scheme:

[0008] A full coverage cleaning path planning method based on Bezier curve optimization comprises the following steps:

[0009] Step S1, region preprocessing module: obtain the grid map and polygon description of the cleaning region, perform map cost conversion, polygon clipping and rotation correction, and extract the free space cell set in the region; the rotation correction is based on the minimum circumscribed rectangle of the region to determine the rotation angle, so that the longest side of the region is parallel to the coordinate axis;

[0010] Step S2, path generation module: according to the free space cell set and the region characteristics, adaptively select the "zigzag" or "back" algorithm to generate a basic full coverage path, and the basic full coverage path comprises an initial polyline path of multiple path points;

[0011] Step S3, path optimization module: based on the basic full coverage path, the Bezier curve is used for smoothing processing to generate a smooth path; wherein the smoothing processing comprises: extracting key points from the basic full coverage path as control points of the Bezier curve; according to the number of control points, the order of the Bezier curve is dynamically selected; based on the Bezier curve of the selected order and the control points, a smooth continuous curve path is generated;

[0012] Step S4, path adjustment module: the optimized path is inversely transformed to the original coordinate system, breakpoint connection, abnormal point filtering and boundary path supplement are performed, and a final executable smooth full coverage path is generated.

[0013] Preferably, in step S1, the specific steps of the rotation correction comprise:

[0014] The minimum circumscribed rectangle of the cleaning region polygon and its direction angle are calculated;

[0015] The rotation angle θ is determined according to the height-width ratio of the minimum circumscribed rectangle, and the determination of the rotation angle θ is as follows: if the height of the minimum circumscribed rectangle is greater than the width,

[0016] then, θ = -(90+rect.angle) / 180.0 x π;

[0017] Otherwise, θ = -rect.angle / 180.0 x π;

[0018] Converting the grid map to a cost map and rotating the polygon vertices by an angle θ with the center point of the polygon as the rotation center.

[0019] Preferably, in the step S2:

[0020] The "zigzag" path generation is a search in the horizontal direction, and the next passing cell is selected by combining the cell passing cost and the weight function of direction continuity. The "zigzag" path is suitable for cleaning an open area.

[0021] The "back" path generation is a multi-directional search along the polygon boundary and the obstacle edge to generate a surrounding path. The "back" path is suitable for cleaning a complex area with obstacles.

[0022] Preferably, in the step S3, the vertices of the base path are screened, the starting point, the ending point, the intermediate key turning points (points with a turning angle greater than a set threshold), and the region midpoint are retained, the redundant vertices are removed, the control point sequence of the Bezier curve is generated, and the order of the Bezier curve is dynamically selected according to the number of the screened control points:

[0023] When the number of control points is 2, a first-order Bezier curve (straight line fitting) is selected;

[0024] When the number of control points is 3, a second-order Bezier curve is selected;

[0025] When the number of control points is 4, a third-order Bezier curve is selected,

[0026] When the number of control points is greater than or equal to 5, a fourth-order Bezier curve is selected.

[0027] Preferably, in the step S3, the introduced geometric center point is the geometric center of the cleaning area polygon.

[0028] Preferably, the step S3 further includes discretely sampling the Bezier curve and calculating the continuous motion direction angle of each sampling point.

[0029] Preferably, the step S3 further includes:

[0030] For the discretely sampled path points, the direction angle θ = atan2(Δy, Δx) of adjacent two points is calculated.

[0031] The direction angle is converted to a quaternion form: w = cos(θ / 2), z = sin(θ / 2), to ensure the continuity of the robot movement direction.

[0032] Preferably, in step S4, the boundary path supplementation includes:

[0033] Generate a path that fits the polygonal boundary of the cleaning area;

[0034] After smoothing using Bézier curves, the boundary path is optimized.

[0035] Seamlessly connect the boundary path with the main path.

[0036] In summary, this application includes at least one of the following beneficial effects:

[0037] 1. This application creatively improves the contradiction between "path smoothness" and "full coverage rigor" by combining regional adaptive preprocessing, targeted basic path generation and constrained Bézier curve optimization. In step S3, instead of blindly smoothing the entire path globally, key inflection points are selected and regional geometric center points are introduced as control points. This allows the generated Bézier curve to prioritize strict coverage of key areas such as regional boundaries and corners while maintaining overall smoothness. Then, through boundary path supplementation and adjustment in step S4, edge "gaps" or "overflows" that may occur during the smoothing process are further eliminated. Thus, the smoothness of the path is greatly improved without sacrificing the coverage integrity.

[0038] 2. This application also achieves adaptation to cleaning scenarios of different shapes and complexities through the rotation correction preprocessing in step S1 and the adaptive path generation strategy in step S2. The rotation correction is performed according to the direction of the minimum bounding rectangle of the area, laying the foundation for the subsequent generation of regular paths. In addition, it can adaptively select the "zigzag" or "return" basic path algorithm according to the characteristics of the area (such as whether it contains obstacles or whether the shape is regular). This enables the generation of efficient round-trip paths in open areas and the generation of boundary-fitting loop paths in complex obstacle areas. Thus, it can ensure high coverage and planning rationality in different scenarios, improving the versatility of the method and the actual cleaning efficiency. Attached Figure Description

[0039] Figure 1 This is a flowchart of the region preprocessing module in this embodiment of the application;

[0040] Figure 2 This is a schematic diagram of the Bezier curve optimization in this embodiment of the application. Detailed Implementation

[0041] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0042] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0043] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0044] This application discloses a full-coverage cleaning path planning method based on Bézier curve optimization, implemented in a robot operating system (such as ROS, ROS2) environment. The robot is equipped with environmental perception sensors (such as LiDAR and vision sensors) to acquire environmental grid map information in real time. This method achieves efficient, smooth, and spot-free cleaning path planning by performing rotational correction on the cleaning area, generating a basic full-coverage path, optimizing path smoothness using Bézier curves, and adjusting path direction and boundary adaptation.

[0045] This method mainly includes four core processing modules, which are listed in order:

[0046] The regional preprocessing module is responsible for standardizing the input environmental information and preparing data for path generation.

[0047] Path generation module: Adaptively generates basic, fully covered paths based on regional features;

[0048] Path optimization module: Uses Bézier curves to smooth and optimize the basic path;

[0049] Path adjustment module: Adjusts the optimized path for compatibility and generates the final executable path.

[0050] Data is passed sequentially between modules, with the output of one module serving as the input of the next.

[0051] The specific implementation steps are as follows:

[0052] (1) Step S1, Area Preprocessing Module: This module is used to obtain the raster map and polygon description of the cleaning area, perform map cost conversion, polygon clipping and rotation correction, and extract the set of free space cells within the area. Among them, the rotation correction is based on the minimum bounding rectangle of the area to determine the rotation angle, so that the longest side of the area is parallel to the coordinate axis.

[0053] The region preprocessing module operates according to the following steps:

[0054] Raster map to cost map conversion: The system receives raster map data in the standard format of navigation messages (nav_msgs::OccupancyGrid) from the Robot Operating System 2 (ROS2). Each raster value (range 0-100, representing the occupancy probability) is mapped to a cost value (range 0-254). The cost value for unknown areas is set to NO_INFORMATION, the cost value for identified obstacle areas is set to LETHAL_OBSTACLE, and the cost value for free space is calculated linearly based on its original raster values. Finally, a standard two-dimensional cost map (costmap_2d::Costmap2D) in the ROS2 operating system is generated.

[0055] Polygon region clipping: Process the input cleaned region polygon path (nav_msgs::Path type). If the distance between the first and last endpoints of the polygon is less than the set threshold (CHECK_DIST) and the number of vertices exceeds the set value (CHECK_MAX), clip redundant vertices and remove duplicate endpoints to ensure that the final polygon region is closed and has no redundant vertices.

[0056] Region rotation correction: The longest side direction of the polygon region is calculated using the minimum bounding rectangle algorithm. The cost map and the polygon region are then rotated along this longest side, aligning it with the coordinate axes and reducing the number of path turns. During rotation, the region's center point is used as the rotation center, and the map and polygon vertices are rotated using coordinate transformation formulas. The rotation angle θ is calculated using the angle of the minimum bounding rectangle.

[0057] If the height of the smallest bounding rectangle is greater than its width:

[0058] Then θ = -(90 + rect.angle) / 180.0 × π;

[0059] Otherwise, θ = -rect.angle / 180.0×π;

[0060] Where rect.angle is the angle between the rectangle and the horizontal axis.

[0061] Using the center of the polygonal region as the center of rotation, apply the coordinate transformation formula:

[0062] x'=(x-cx)*cosθ-(y-cy)*sinθ+cx;

[0063] y'=(x-cx)*sinθ+(y-cy)*cosθ+cy;

[0064] Rotate the entire free space reference frame of the cost map and all vertices of the polygon. This operation aligns the longest side of the region with the X-axis of the coordinate system. If the height is greater than the width, calculate θ according to the formula to align the longest side (in this case, the height) with the Y-axis, which is logically equivalent.

[0065] Free Space Extraction: Based on the cost map, free space cells within the polygonal region are extracted using the FreeSpaceCalculator class. A ray casting method is used to determine if each cell is inside the polygon. For cells inside the polygon, their cost value on the cost map is checked to see if it is lower than a preset free space threshold (cell_free_thres_), for example, 60. Cells that simultaneously meet the criteria of being "inside the polygon" and having a cost value below the threshold are marked as cleanable "free space cells." For each such cell, the module records an "optimal access point": if the cell is inside free space, the access point is selected as the cell center; if it is near a boundary or obstacle, a more internal location is selected based on proximity information to reduce collision risk. Finally, an ordered or indexable set of free space cells is obtained.

[0066] (2) Step S2, Path generation module, according to the cleaning scene requirements, adaptively selects the "zigzag" or "backshaped" path generation algorithm, and generates a basic full-coverage path based on the preprocessed free space cells. The basic full-coverage path contains an initial polyline path with multiple path points.

[0067] The zigzag path generation includes:

[0068] Determine the starting point: Select the smallest vertex of the polygon region boundary and map it to the nearest free space cell in the cost map as the starting point. The initial orientation is set according to the position of the starting point (180° if there is an obstacle or boundary on the right, otherwise 0°).

[0069] Path search: Based on the weight guidance of the neural matrix (neural_mat_), the next passable cell is searched in the horizontal direction (0° and 180°). The weight calculation combines the cell cost value and the direction continuity factor (e = 1 - |Δθ| / 180, where Δθ is the angle between the current orientation and the target direction), and the direction with the largest weight is selected first.

[0070] Breakpoint handling: If there are no feasible cells in the current direction, calculate the distance between the current position and all free space cells, select the nearest feasible cell as the next target point, and generate a breakpoint connection path.

[0071] The generation of the zigzag path includes:

[0072] Starting point determination: Consistent with the alphanumeric path, the starting orientation is set according to the clockwise / counterclockwise characteristics of the polygonal region (clockwise orientation is 180°, counterclockwise orientation is 0°).

[0073] Direction Search: Define 8 search directions (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°). Based on the current orientation, rotate the search direction sequence and prioritize searching feasible cells in directions near the current orientation.

[0074] Boundary adaptation: Generate a surrounding path along the polygon boundary and obstacle edge to ensure that there are no missed areas around the obstacle.

[0075] For example, in this embodiment, the cleaning scenario is a living room. Since there are multiple obstacles such as an L-shaped sofa and a coffee table, it is a complex area. Therefore, the "backshaped" path generation algorithm is selected to achieve better boundary and obstacle adaptation.

[0076] Generating a U-shaped path for a living room scene:

[0077] a. Determining the starting point: In the set of free space cells, find the boundary point of the polygon (after rotation) that is smallest in both the X and Y directions, and map it to the center of the nearest free space cell, setting it as the starting point of the path. Based on the order of the polygon vertices, it is determined to be clockwise, so the initial orientation is set to 180 degrees (facing west).

[0078] b. Direction Search and Path Extension: Define 8 search directions (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°). Starting from the initial point, using the current orientation (180°) as a baseline, prioritize trying adjacent cells in the current orientation and its left and right neighboring directions (e.g., 135°, 180°, 225°). Obtain the weights of candidate directions by querying the neural matrix `neural_mat_`. This weight combines the cost value of the target cell (lower cost, higher weight) and the direction change penalty factor e = 1 - |Δθ| / 180 (smaller change, higher weight). Select the feasible direction with the highest weight as the next step.

[0079] Boundary and Obstacle Encirclement: The algorithm first guides the path along the outer boundary of the polygon, forming the outer contour. When encountering an internal obstacle (such as a coffee table), the algorithm naturally avoids it during weight calculation because the cost of cells around the obstacle increases sharply, thus guiding the path to follow the edge of the obstacle, forming an inner encirclement. After a closed loop is completed, the algorithm adjusts the priority of the search direction to guide the path to contract inward, starting the coverage of the next loop, forming a spiral "U" shaped structure.

[0080] c. Breakpoint Handling: A breakpoint is identified when no matching free space cells (i.e., surrounded by covered areas or obstacles) are found at a given location in all preset search directions. The distances from the current location to all uncovered free space cells are calculated, the nearest cell is selected as the next target point, and a straight line or simple polyline is planned to connect it, avoiding obstacles. The search continues from this point in a zigzag pattern. Ultimately, a basic path covering all free space cells is generated. This path consists of a series of continuous coordinate points, but exhibits a distinct polyline at corners.

[0081] (3) Step S3, Path Optimization Module: Based on the basic full-coverage path, a smooth path is generated by using a Bézier curve for smoothing. The smoothing process includes: extracting key points from the basic full-coverage path as control points for the Bézier curve; dynamically selecting the order of the Bézier curve based on the number of control points; and generating a smooth, continuous curve path based on the Bézier curve of the order and the control points.

[0082] The path optimization module works in the following steps:

[0083] Control point filtering: Vertices of all basic paths are filtered, retaining only the start and end points. For intermediate points, the direction vectors of the path segments before and after them are calculated. If the angle of change in direction (calculated using dot product and cross product) is greater than a set threshold (e.g., 15 degrees), the point is identified as a "critical inflection point" and retained. Simultaneously, to maintain the path's fit to the overall region, the geometric center points of the polygonal region are also added to the control point sequence. Then, redundant vertices are removed, generating the control point sequence (std::vectorBezier::Point) for the Bézier curve.

[0084] Bézier curve order selection: The curve order is dynamically selected based on the number of control points to ensure curve smoothness and fitting accuracy.

[0085] When the number of control points is 2, select a first-order Bézier curve (linear fitting);

[0086] When the number of control points is 3, select a second-order Bézier curve;

[0087] When the number of control points is 4, select a 3rd order Bézier curve;

[0088] When the number of control points is greater than or equal to 5, select a 4th-order Bézier curve.

[0089] In this embodiment, after screening, 5 control points are obtained (starting point, regional center point, two key inflection points around the sofa, and ending point). The definition is selected according to the order (when the number of control points is ≥5, a 4th order Bézier curve is selected). Therefore, the path adjustment module uses a 4th order Bézier curve for fitting.

[0090] Path discretization: Based on the set step size (step_in_fine_path_), for example 0.03 meters, the Bézier curve is sampled from 0 to 1 on parameter t with an adaptive step size to calculate a series of discrete points, generating a smooth fine path. The number of discrete points is max(curve length / step size, 2) to ensure uniform path density.

[0091] Orientation angle calculation: For each discretized path point, calculate the orientation angle θ = atan2(Δy, Δx) between adjacent points, and convert this orientation angle into the quaternion format commonly used in ROS path messages (w = cos(θ / 2), z = sin(θ / 2)). This generates a smooth path composed of dense points with continuously changing orientation, ensuring continuous robot movement direction.

[0092] (4) Step S4, Path Adjustment Module: Transform the optimized path back to the original coordinate system, connect breakpoints, filter out abnormal points and supplement boundary paths to generate the final executable smooth full-coverage path so that it can be adapted to the actual robot operation.

[0093] The path adjustment module works according to the following steps:

[0094] Boundary adjustment: Rotate the path after rotation correction back to the original coordinate system (i.e. the coordinate system used by the robot to map), and restore the actual position of the path through the inverse coordinate transformation formula, so that the optimized path is smooth in shape and its spatial position is consistent with the boundary of the original cleaning area.

[0095] Breakpoint connection and anomaly filtering: Check for breakpoints in the path. If any two adjacent points are found to be more than 0.1 meters apart (possibly caused by previous breakpoint jumps), linear interpolation is used to fill in the gap between these two points to ensure the physical continuity of the path. Traverse all points and delete any invalid points whose position coordinates and orientation quaternions are both 0 to avoid robot malfunctions.

[0096] Boundary path supplementation: To prevent missed areas due to smoothing at the boundaries, the path adjustment module offsets the original polygon boundary inward by a distance equal to the robot's coverage radius, generating a boundary path that surrounds the cleaning area. The boundary path is lightly smoothed using a low-order Bézier curve (e.g., order 2), and its beginning and end are connected to the start and end points of the main smoothed path to form a complete final cleaning path. This ensures no missed areas at the region edges and a seamless connection between the boundary path and the main path.

[0097] Output: The final processed path point sequence and its direction quaternion are encapsulated into a standard nav_msgs::Path message and published to the robot's underlying motion controller for execution, thereby completing efficient, smooth, and full-coverage cleaning of the specified area.

[0098] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A full coverage cleaning path planning method based on Bezier curve optimization, characterized in that: The method comprises the following steps: Step S1, a region preprocessing module: obtaining a grid map and a polygon description of a cleaning region, performing map cost conversion, polygon clipping and rotation correction, and extracting a free space cell set in the region; the rotation correction is based on a minimum circumscribed rectangle of the region to determine a rotation angle, so that the longest side of the region is parallel to the coordinate axis; Step S2, a path generation module: adaptively selecting a "zigzag" or "back-to-back" algorithm according to the free space cell set and the region characteristics, generating a basic full-coverage path, and the basic full-coverage path comprises an initial polyline path of multiple path points; Step S3, a path optimization module: based on the basic full-coverage path, performing smoothing processing by using a Bezier curve to generate a smooth path; wherein the smoothing processing comprises: extracting key points from the basic full-coverage path as control points of the Bezier curve; dynamically selecting the order of the Bezier curve according to the number of control points; based on the Bezier curve of the selected order and the control points, a smooth continuous curve path is generated; Step S4, a path adjustment module: transforming the optimized path to the original coordinate system, performing breakpoint connection, abnormal point filtering and boundary path supplement, and generating a final executable smooth full-coverage path.

2. The full coverage cleaning path planning method based on Bezier curve optimization according to claim 1, characterized in that: In the step S1, the specific steps of the rotation correction include: calculating the minimum circumscribed rectangle of the polygon of the cleaning region and its direction angle; determining the rotation angle θ according to the height-width ratio of the minimum circumscribed rectangle, wherein if the height of the minimum circumscribed rectangle is greater than the width, then θ = -(90 + rect.angle) / 180.0 * π; otherwise, θ = -rect.angle / 180.0 * π; rotating the grid map and the polygon vertices by θ degrees with the center point of the polygon as the rotation center.

3. The full coverage cleaning path planning method based on Bezier curve optimization according to claim 1, characterized in that: In the step S2: the "zigzag" path generation is to search in the horizontal direction, and the next passing cell is selected by combining the cell passing cost and the weight function of direction continuity, and the "zigzag" path is suitable for cleaning an open region; the "back-to-back" path generation is to generate a surrounding path along the polygon boundary and the obstacle edge by multi-directional search, and the "back-to-back" path is suitable for cleaning a complex region with obstacles.

4. The full coverage cleaning path planning method based on Bezier curve optimization according to claim 1, characterized in that: In the step S3, the vertices of the basic path are screened, the starting point, the ending point, the intermediate key inflection points (the points where the path direction changes greater than a set threshold) and the region midpoint are retained, the redundant vertices are removed, the control point sequence of the Bezier curve is generated, and the order of the Bezier curve is dynamically selected according to the number of screened control points: when the number of control points is 2, a first-order Bezier curve (straight line fitting) is selected; when the number of control points is 3, a second-order Bezier curve is selected; when the number of control points is 4, a third-order Bezier curve is selected, when the number of control points is greater than or equal to 5, a fourth-order Bezier curve is selected.

5. The full coverage cleaning path planning method based on Bezier curve optimization according to claim 4, characterized in that: In the step S3, the introduced geometric center point is the geometric center of the polygon of the cleaning region.

6. The full coverage cleaning path planning method based on Bezier curve optimization according to claim 3, characterized in that: The step S3 further comprises: discretely sampling the Bezier curve, and calculating the continuous motion direction angle of each sampling point.

7. The full coverage cleaning path planning method based on Bezier curve optimization according to claim 6, characterized in that: The step S3 further comprises: For the discretely sampled path points, the direction angle θ = atan2(Δy, Δx) of adjacent two points is calculated; The direction angle is converted into quaternion form: w = cos(θ / 2), z = sin(θ / 2), to ensure the continuity of the robot moving direction.

8. The full coverage cleaning path planning method based on Bezier curve optimization according to claim 1, characterized in that: In the step S4, the boundary path supplementing comprises: Generating a path adhering to the boundary of the cleaning area polygon; After smoothing by using the Bezier curve, the boundary path is optimized; The boundary path is seamlessly connected with the main path.

Images