Optimization Method for Obstacle Avoidance Path of Cleaning Robots in Multi-Obstacle Scenarios

By constructing a minimum turning envelope area and identifying bottleneck areas, and configuring appropriate operation control parameters, the problem of drive wheel slippage and insufficient braking caused by cleaning fluid deposition in multi-obstacle scenarios was solved, thus improving the safety and efficiency of path planning.

CN123113885APending Publication Date: 2026-11-17BEIJING KAIPU ZHIYUAN TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611461012.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing cleaning robots fail to effectively consider the changes in ground wetness and friction coefficient caused by cleaning fluid deposition in multi-obstacle scenarios, leading to problems such as drive wheel slippage, insufficient braking, or inability to return.

Method used

By collecting obstacle contours and ground material data, a minimum turning envelope region is constructed, bottleneck regions are identified, and operational control parameters such as travel speed, cleaning fluid output, and cleaning pressure are configured for candidate obstacle avoidance paths. The lower bound of the friction coefficient is calculated, and paths that meet the motion constraints are selected.

Benefits of technology

It enables a comprehensive assessment of the traffic capacity and movement safety in bottleneck areas, improves the accuracy of slip risk assessment, shortens operation time, and maintains stable cleaning quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113885A_ABST
    Figure CN123113885A_ABST
Patent Text Reader

Abstract

This invention discloses a method for optimizing obstacle avoidance paths for cleaning robots in multi-obstacle scenarios, relating to the field of robot path planning technology. The method involves collecting obstacle contours, ground material, and the robot's pose; constructing a minimum turning envelope region and identifying bottleneck regions; generating candidate obstacle avoidance paths that pass through bottleneck regions; configuring travel speed, cleaning fluid output, water absorption power, and cleaning pressure for each path; determining wheel track regions, cleaning fluid deposition regions, and arrival times; calculating temporal wettability and obtaining the lower bound of the friction coefficient; calculating the driving force and braking force margins for entering and returning to the bottleneck region; filtering out paths that do not meet the constraints; selecting the target path with the shortest travel time; and controlling the robot to perform cleaning operations. This invention achieves collaborative optimization of path and operational parameters through bottleneck region identification, wettability assessment, and mechanical margin selection, which helps reduce the risk of drive wheel slippage and insufficient braking, improving passage safety and cleaning efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of robot path planning technology, specifically to a method for optimizing obstacle avoidance paths for cleaning robots in multi-obstacle scenarios. Background Technology

[0002] With the development of indoor automatic cleaning equipment, cleaning robots usually obtain the outline of obstacles through sensors, plan obstacle avoidance paths, and simultaneously perform liquid supply, wiping and water suction operations. Existing plans are mainly based on the geometric boundaries of obstacles, robot size and travel distance, and preset operation parameters according to the ground material, with little consideration for the impact of changes in cleaning liquid deposition on the safety of path passage.

[0003] Existing methods generate passable paths based on obstacle contours, and then perform operations at fixed speeds and cleaning parameters. After the cleaning fluid is output, a sedimentation area is formed on the ground. When water absorption and recovery are insufficient or the robot repeatedly passes through the same location, the amount of residual fluid increases, the wettability increases, and the actual friction coefficient of the wheel track area decreases. Existing paths do not dynamically calculate the friction coefficient by combining the sedimentation time, ground material, and operation parameters. In bottleneck areas jointly defined by multiple obstacles, when the robot enters or returns, the driving force or braking force may exceed the wheel-ground adhesion capacity, resulting in drive wheel slippage, deviation from the driving path, collision with obstacles, or inability to return. Summary of the Invention

[0004] The purpose of this invention is to provide a path optimization method for cleaning robots in multi-obstacle scenarios, in order to solve the technical problem mentioned in the background art that the existing obstacle avoidance path planning does not consider the changes in ground wetness and friction coefficient caused by cleaning liquid deposition, which leads to the cleaning robot's drive wheels slipping easily, insufficient braking or inability to return in bottleneck areas formed by multiple obstacles.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing obstacle avoidance paths for cleaning robots in multi-obstacle scenarios, comprising: S1. Collect obstacle contours, ground material and cleaning robot pose. Based on the minimum turning radius, scan the cleaning robot to obtain the minimum turning envelope area. Determine the passable area according to the obstacle contour. Identify the passable area that is jointly limited by multiple obstacles and cannot accommodate the minimum turning envelope area as the bottleneck area. S2. Generate candidate obstacle avoidance paths through the bottleneck area based on the robot's pose. Configure operation control parameters consisting of travel speed, cleaning fluid output, water absorption power, and cleaning pressure for each candidate obstacle avoidance path. Determine the wheel track area, cleaning fluid deposition area, and path arrival sequence based on the candidate obstacle avoidance path and operation control parameters. S3. Calculate the temporal wetness of the cleaning fluid deposition area by means of path arrival time, ground material and operation control parameters, and obtain the lower bound of the friction coefficient of the wheel track area from the pre-established wetness-friction coefficient mapping relationship by means of temporal wetness. S4. Based on the lower bound of the friction coefficient, calculate the driving force margin and braking force margin corresponding to the cleaning robot entering the bottleneck area and returning to the entrance of the bottleneck area along the candidate obstacle avoidance path, and screen out any candidate obstacle avoidance path whose mechanical margin does not meet the motion constraints. S5. Select the target obstacle avoidance path with the shortest travel time from the retained candidate obstacle avoidance paths, and control the cleaning robot to perform cleaning operations according to the target obstacle avoidance path and the corresponding operation control parameters.

[0006] Preferably, the specific process for obtaining the minimum turning envelope region is as follows: Obtain the initial projection envelope of the cleaning robot in the ground coordinate system. Establish a turning trajectory from the initial heading to the reverse heading based on the minimum turning radius. Transform the center coordinates and heading angle of the initial projection envelope along the turning trajectory. Take the union of all transformed projection envelopes to obtain the minimum turning envelope region.

[0007] Preferably, the specific process for identifying the bottleneck region is as follows: The obstacle contours are mapped onto the ground coordinate system. The area enclosed by the obstacle contours is removed from the working area to obtain the passable area. The projection envelope of the cleaning robot is moved within the passable area. The positions where the projection envelope does not intersect with the obstacle contours form a set of straight-line passable positions. The minimum turning envelope area is moved within the set of straight-line passable positions. The positions where the minimum turning envelope area intersects with the obstacle contours form a set of turning-restricted positions. The continuous position area in the set of turning-restricted positions, which is jointly defined by multiple obstacles, is determined as the bottleneck area.

[0008] Preferably, the specific process of generating candidate obstacle avoidance paths through the bottleneck region is as follows: Extract the boundary sections at both ends of the bottleneck region, determine the center of the boundary section of the adjacent cleaning robot pose as the bottleneck inlet, and determine the center of the other boundary section as the bottleneck outlet. Generate the bottleneck centerline based on the midpoint of each cross section between the bottleneck inlet and the bottleneck outlet. Generate offset paths that satisfy the projection envelope passage constraints on both sides of the bottleneck centerline. Generate an approach path connecting the bottleneck inlet based on the position coordinates and heading angle of the cleaning robot pose. Concatenate the approach path with the bottleneck centerline and the offset path respectively to obtain the candidate obstacle avoidance path.

[0009] Preferably, the specific process for determining the wheel track region, the cleaning fluid deposition region, and the path arrival time sequence is as follows: The cleaning robot includes drive wheels and cleaning components. The candidate obstacle avoidance path is divided into multiple path segments according to the ground material. The travel speed, cleaning fluid output, water absorption power and cleaning pressure are configured for each path segment to form a sequence of operation control parameters. The ground contour of the drive wheels is scanned along each path segment to obtain the wheel track area. The ground contour of the cleaning components is scanned along each path segment to obtain the cleaning fluid deposition area. The travel time of each path segment is calculated based on the length and travel speed of each path segment. The travel time of each path segment is accumulated according to the connection order of the path segments to obtain the path arrival time sequence.

[0010] Preferably, the specific process for calculating the temporal wettability of the cleaning fluid deposition area is as follows: The liquid supply per unit area is calculated based on the cleaning fluid output, travel speed, and grounding width of the cleaning components. The liquid recovery per unit area is determined based on the water absorption power. The difference between the liquid supply per unit area and the liquid recovery per unit area is determined as the residual liquid per unit area. The residual liquid per unit area is converted into initial wettability based on the cleaning pressure and the liquid absorption parameters corresponding to the ground material. The duration after each cleaning fluid deposition is determined based on the path arrival time sequence. The initial wettability is attenuated using the wettability attenuation parameters corresponding to the ground material. The attenuation results of the repeated traversal locations are added together to obtain the time-series wettability.

[0011] Preferably, the specific process of pre-establishing the wettability-friction coefficient mapping relationship is as follows: Multiple humidity sampling states are set for each type of ground material. Under constant normal load, the output torque of the drive wheel is increased step by step. The critical traction force when the drive wheel reaches the slip threshold is recorded. The friction coefficient is obtained by calculating the ratio of the critical traction force to the constant normal load. The humidity sampling states and the friction coefficient are combined to form a calibration sample. The calibration sample is monotonically fitted according to the direction of increasing humidity. The friction coefficient confidence interval is generated based on the fitting error. The lower boundary of the friction coefficient confidence interval is determined as the humidity-friction coefficient mapping relationship corresponding to the ground material.

[0012] Preferably, the specific process for calculating the driving force margin and braking force margin is as follows: The upper limit of adhesion force is calculated based on the lower limit of the friction coefficient and the normal load of the drive wheel. The required driving force is calculated based on the changes in travel speed, rolling resistance, and cleaning resistance generated by cleaning pressure. The driving force margin is obtained by subtracting the required driving force from the upper limit of adhesion force. The required braking force is calculated based on the travel speed and the remaining braking distance in the bottleneck area. The braking force margin is obtained by subtracting the required braking force from the upper limit of adhesion force. The driving force margin and braking force margin corresponding to entering the bottleneck area and returning to the bottleneck area entrance are calculated separately according to the forward path segment order and the reverse path segment order of the candidate obstacle avoidance path.

[0013] Preferably, the process for selecting the target obstacle avoidance path is as follows: The driving force margin and braking force margin are compared with their respective preset safety margin thresholds. Candidate obstacle avoidance paths with all mechanical margins not less than the safety margin threshold are retained. The travel time of each retained candidate obstacle avoidance path segment and the switching time of operation control parameters are calculated. The sum of the travel time of the path segment and the switching time of operation control parameters is determined as the passage time. The candidate obstacle avoidance path with the shortest passage time is selected as the target obstacle avoidance path.

[0014] Preferably, each path segment of the target obstacle avoidance path corresponds to a corresponding operation control parameter. The corresponding operation control parameter is determined as the target operation control parameter. The current path segment is determined according to the pose of the cleaning robot. The traveling speed, cleaning liquid output, water absorption power and cleaning pressure corresponding to the current path segment are read from the target operation control parameter to control the cleaning robot to perform cleaning operations.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention collects obstacle contours and constructs a minimum turning envelope region based on the minimum turning radius of the cleaning robot. It identifies continuous passable areas that are jointly defined by multiple obstacles and cannot be contained within the envelope region as bottleneck areas. Then, it filters candidate obstacle avoidance paths based on the driving force margin and braking force margin in the entry and return directions. This enables a comprehensive judgment on bottleneck passability and movement safety, and solves the problem that planning paths based solely on obstacle spacing cannot identify drive wheel slippage and insufficient braking in narrow areas. This helps reduce the risk of collisions and inability to return. 2. This invention determines the wheel track area, cleaning fluid deposition area, and path arrival sequence by configuring the travel speed, cleaning fluid output, water absorption power, and cleaning pressure for candidate obstacle avoidance paths. It also calculates the temporal wetness based on the ground material and deposition duration, and obtains the lower bound of the friction coefficient through the wetness-friction coefficient mapping relationship. This enables dynamic evaluation of wheel-ground adhesion changes caused by cleaning operations, solves the problem of overestimation of adhesion ability and distortion of safety screening due to failure to consider residual liquid accumulation, and helps improve the accuracy of slip risk assessment. 3. This invention selects the target obstacle avoidance path with the shortest passage time by adding the travel time of each path segment of the candidate obstacle avoidance path to the switching time of the operation control parameters. It also reads the travel speed, cleaning fluid output, water absorption power and cleaning pressure of the path segment corresponding to the cleaning robot's real-time pose, thus realizing the synchronous execution of the path selection result and the cleaning operation control parameters. This solves the problem of difficulty in balancing passage efficiency, parameter switching delay and on-site execution consistency when multiple paths meet the attachment safety conditions, which is beneficial to shortening the operation time and maintaining stable cleaning quality. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of the present invention; Figure 2 This is a flowchart of the minimum turning envelope and bottleneck region identification process of the present invention; Figure 3 This is a flowchart illustrating the process of generating candidate obstacle avoidance paths that traverse bottleneck regions according to the present invention. Figure 4 This is a flowchart of the screening process for time-series wettability, coefficient of friction, and mechanical margin in this invention. Figure 5 This is a mapping diagram of wettability-friction coefficient according to the present invention. Detailed Implementation

[0017] Please see Figure 1 The present invention provides an embodiment of a path optimization method for obstacle avoidance in cleaning robots in multi-obstacle scenarios, comprising: S1. Before or during the robot begins cleaning, the robot collects the obstacle contours, ground materials, and robot pose within the work area. The robot pose includes its current position and heading. The obstacle contours represent the boundaries of inaccessible areas such as walls, cabinets, and equipment bases. The ground materials distinguish between surfaces with different frictional properties, such as ceramic tiles and epoxy flooring, under the same liquid conditions. The robot pose determines the planning starting point. Based on the robot's dimensions and minimum turning radius, the robot's projection envelope moves continuously along a turning trajectory from the initial heading to the reverse heading. The entire ground area occupied during the movement is merged into the minimum turning envelope area. Then, based on the obstacle contours, the passable area is obtained. The continuous area that is held by multiple obstacles, allowing the robot to maintain its current heading and move straight but cannot accommodate the minimum turning envelope area, is identified as the bottleneck area. This distinguishes between ordinary open areas and narrow passages where the robot cannot turn around on the spot. S2. Starting from the robot's pose, generate multiple candidate obstacle avoidance paths that connect the entrance to the bottleneck area, pass through the interior of the bottleneck area, and return to the entrance of the bottleneck area. Each candidate obstacle avoidance path is divided into path segments according to the ground material and movement stage. The travel speed, cleaning fluid output, water absorption power, and cleaning pressure are configured for each path segment. Based on the installation position of the robot's drive wheels and cleaning components relative to the robot body, scan along each candidate obstacle avoidance path to obtain the wheel track area that the drive wheels may contact, the cleaning fluid deposition area that the cleaning fluid may cover, and the order in which the robot and cleaning components arrive at each position, so that a one-to-one correspondence is established between the geometric path and the cleaning operation status. S3. For each candidate obstacle avoidance path, the degree of cleaning fluid application is determined based on the travel speed and the amount of cleaning fluid output, and the degree of liquid recovery is determined based on the water absorption power. Then, combined with the cleaning pressure, ground material, and the time elapsed after liquid deposition, the wettability of the cleaning fluid deposition area over time is obtained. For the location that is repeatedly cleaned, the wettability is calculated and superimposed according to the deposition time of each deposition. The time-series wettability corresponding to the arrival time of the wheel track is input into the pre-calibrated wettability-friction coefficient mapping relationship to obtain the lower bound of the friction coefficient that can be conservatively adopted for the wheel track area at the corresponding time, thereby avoiding overestimation of the adhesion ability by only using the friction coefficient of dry ground. S4. The available adhesion force is determined by the lower bound of the friction coefficient and the normal load of the drive wheel. Combined with the speed change, rolling resistance, cleaning resistance caused by cleaning pressure, current speed and remaining braking distance, the driving force margin and braking force margin when the robot enters the bottleneck area along the candidate obstacle avoidance path and returns to the bottleneck entrance in the opposite direction are obtained respectively. As long as any mechanical margin in any direction and any path position is lower than the corresponding motion constraint, the candidate obstacle avoidance path is screened out. This can eliminate the possibility of drive wheel slippage or failure to stop within a limited distance in wet narrow channels in advance. S5. For the remaining candidate obstacle avoidance paths, the travel time and operation control parameter switching time of each path segment are accumulated, and the path with the shortest passage time is selected as the target obstacle avoidance path. The controller identifies the current path segment of the robot based on the real-time pose, reads the corresponding travel speed, cleaning fluid output, water absorption power and cleaning pressure of the path segment and controls the corresponding actuators. In the bottleneck formed by multiple obstacles, passability, wet ground adhesion conditions, braking safety and cleaning efficiency are all taken into account in the path selection, so that the robot can not only pass through narrow areas, but also complete the cleaning operation according to the mechanically verified shorter time plan.

[0018] Please see Figure 1 and Figure 2 The present invention provides an embodiment of a cleaning robot obstacle avoidance path optimization method for multi-obstacle scenarios, which collects obstacle contours, ground materials and cleaning robot poses, scans the cleaning robot based on the minimum turning radius to obtain the minimum turning envelope area, determines the passable area based on the obstacle contours, and identifies the passable area that is jointly limited by multiple obstacles and cannot accommodate the minimum turning envelope area as the bottleneck area. Furthermore, obstacle contours, ground materials, and robot poses are represented using a unified ground coordinate system. The robot obtains distance points in the same height plane using a two-dimensional LiDAR and supplements the depth points of low obstacles using a depth camera. The controller projects each point onto the ground coordinate system using pre-calibrated sensor extrinsic parameters, clusters adjacent points, and extracts the outer boundaries of each point cluster to obtain obstacle contours with obstacle numbers. The contours of fixed walls and cabinets can also be obtained from the site map. The real-time detection contours and map contours are merged according to timestamps and coordinate positions. The ground material is classified after the texture and reflection features are collected by the camera facing the ground, or read from the pre-labeled site semantic map, and the classification results are written into the corresponding ground grid. The robot pose is obtained by fusing the results of wheel odometry, inertial measurement unit, and laser synchronous positioning. The position coordinates and heading angles use the timestamp of the acquisition time, and the data from different sensors are aligned to the same time through linear interpolation. To obtain the minimum turning envelope region, first obtain the projected contours of the robot body, drive wheels, cleaning components, and extended parts on the ground from the robot structure diagram and physical measurement results, and then extend them outward by a safe distance to form the initial projected envelope P0, with the minimum turning radius R. min The maximum value between the theoretical value of the steering mechanism and the low-speed actual vehicle turning test value is taken. The actual vehicle turning test value is obtained by the robot driving at low speed with the maximum allowable turning angle, recording the positioning trajectory and fitting the trajectory circle. Using the maximum value can take factors such as tire side deviation and assembly error into the turning constraint. In a unified ground coordinate system, with the lower left corner of the working area as the origin, the x-axis and y-axis are mutually perpendicular horizontal coordinate axes. The unit of position coordinates is meters, and the unit of heading angle is radians. Let the center coordinates of the robot's initial projection envelope be: c0 = [x0, y0]; Where x0 is the x-coordinate of the initial center and y0 is the y-coordinate of the initial center, and the robot's initial heading angle is φ0, and θ is the rotation angle that has been turned relative to the initial heading, then the center rotation trajectory from the initial heading to the reverse heading can be expressed as: x(θ)=x0+R min ×[sin(φ0+θ)-sin(φ0)]; y(θ) = y0 - R min ×[cos(φ0+θ)-cos(φ0)]; 0≤θ≤π; Where x(θ) represents the ground x-coordinate of the robot's projection envelope center at the rotation angle θ, y(θ) represents the ground y-coordinate of the center at the rotation angle θ, and x0 and y0 represent the ground x-coordinate and ground y-coordinate of the robot's projection envelope center at the start of rotation, respectively. R minφ0 represents the minimum turning radius after taking into account tire side slip and assembly error, φ0 represents the initial heading angle of the robot relative to the ground coordinate system at the start of the turn, and θ represents the cumulative rotation angle of the robot relative to the initial heading. When θ=0, x(θ)=x0, y(θ)=y0; This indicates that the turning trajectory starts from the initial center. When θ=π, the robot's heading changes 180 degrees relative to the initial heading, reaching the reverse heading. The center coordinates and heading angle of the initial projection envelope are simultaneously transformed along the central turning trajectory. To illustrate the coordinate transformation, the following definition is used: c(θ)=[x(θ),y(θ)]; c(θ) represents the ground coordinates of the robot's projection envelope center at a rotation angle θ. P0 is defined as the set of relative center points obtained by subtracting c0 from the coordinates of each point in P0, with the initial center c0 as a reference. After rotating by an angle θ relative to the initial heading, the transformed projection envelope is: P(θ)=c(θ)+Rot(θ)×[P0-c0]; Where P(θ) represents the set of points in the ground coordinate system whose projection envelope is at a rotation angle θ, and Rot(θ) is a two-dimensional rotation matrix relative to the initial heading with a rotation angle θ, specifically: Rot(θ)=[(cosθ,-sinθ), (sinθ, cosθ)]; The top left, top right, bottom left, and bottom right elements of the matrix represent (cosθ), (-sinθ), (sinθ), and (cosθ), respectively. Matrix multiplication rotates each relative center point in P0-c0 by an angle θ. Since P0 is already placed according to φ0, Rot(θ) only represents the incremental rotation relative to the initial heading and does not multiply by φ0 again. c(θ) is responsible for determining the translation position of the envelope center, and Rot(θ) is responsible for determining the rotation attitude of the envelope shape relative to the initial heading. By taking the union of the transformed projection envelopes corresponding to all turning angles, the minimum turning envelope region is obtained: E turn =∪P(θ), 0≤θ≤π; Among them, E turn The minimum turning envelope region is represented by the symbol ∪, which represents the union of all P(θ) within the range of θ values ​​from 0 to π. This represents the complete turning process from the initial heading to the reverse heading. During calculation, θ can be sampled at angular intervals no greater than two degrees, and the gaps between adjacent envelopes are filled using the union method. The angular interval is determined by offline convergence experiments. When the change in the envelope boundary caused by further reducing the interval is less than the map resolution, the current interval is adopted. When identifying bottleneck regions, first denote the region enclosed by the contour of the i-th obstacle as B.i Where i is the obstacle number, and the work area is denoted as A. work A work If the area represents the entire ground area where the robot needs to perform cleaning tasks, then the passable area is: A free =A work -∪B i ; Among them, A free ∪B represents the passable area after removing all obstacles from the work area. i This represents finding the union of all obstacle regions, and further dividing A... free Discretize the map into a grid, with the grid center p as the candidate robot center, where p is a ground coordinate point. Place the initial projection envelope according to the robot's detected heading at that location. If the projection envelope does not intersect with any obstacle, then that location is added to the set of straight-ahead passable locations C. line If the minimum turning envelope region is placed at the same location and then intersects with an obstacle, then that location is simultaneously added to the set of turning-restricted locations C. lim The corresponding judgment conditions are: C line ={p|P(p)∩B i =Φ, which holds true for all i; C lim ={p|p∈C line And E turn (p)∩B i ≠Φ, holds true for at least one i; Where P(p) represents the projected envelope after placing the center of the initial projected envelope at the candidate center p and rotating it according to the candidate heading, the symbol ∩ represents the intersection of sets, and Φ represents the empty set. Therefore, P(p) ∩ B i =Φ indicates that the projection envelope does not overlap with the i-th obstacle, E turn (p) indicates that E turn The minimum turning envelope region after the reference center is translated to the candidate center p, p∈C line This indicates that candidate center p first meets the straight-through passage condition, E turn (p)∩B i ≠Φ indicates that the rotation envelope overlaps with the i-th obstacle; For C lim Perform connected component analysis and record the obstacle numbers that restrict each location. If the two sides of a continuous location region are formed by two or more obstacles with different numbers, and the initial projection envelope of the region can be passed straight through and the minimum turning envelope region cannot be completely included, then the continuous location region is identified as a bottleneck region. For example, in an indoor work area 8 meters long and 6 meters wide, LiDAR and site map identify cabinet B1 and wall B2. The minimum distance between the relative boundaries of B1 and B2 is 0.86 meters. The ground camera identifies the space within this gap as tile. The robot fuses its positioning pose into position coordinates (1.20 meters, 0.80 meters) and heading angle of zero degrees. The measured projected length of the robot body is 0.72 meters and the width is 0.56 meters. After extending a safe distance of 0.02 meters in all directions, the initial projected envelope is 0.76 meters long and 0.60 meters wide, thus enabling it to move straight along the gap. The theoretical minimum turning radius of the steering mechanism is 0.62 meters. The fitting results of three low-speed turning tests are 0.64 meters, 0.65 meters, and 0.64 meters, respectively. Therefore, R is taken as... min With a minimum turning radius of 0.65 meters, after scanning the initial projection envelope at two-degree intervals, the maximum lateral span of the minimum turning envelope region is 1.84 meters. The maximum lateral span is obtained from the robot size, the 0.65-meter turning radius, and the scan, and is significantly larger than the 0.86-meter interval. Connectivity analysis using a 0.02-meter grid yields a continuous turning-restricted region of 2.40 meters in length, jointly defined by B1 and B2. Therefore, this region is identified as a bottleneck region, indicating that the gap allows the robot to enter straight, but does not allow the robot to make a reverse turn within it.

[0019] Please see Figure 1 and Figure 3 The present invention provides an embodiment of a cleaning robot obstacle avoidance path optimization method for multi-obstacle scenarios, which generates candidate obstacle avoidance paths through bottleneck areas based on the cleaning robot's pose, configures operation control parameters consisting of travel speed, cleaning fluid output, water absorption power and cleaning pressure for each candidate obstacle avoidance path, and determines wheel track area, cleaning fluid deposition area and path arrival sequence based on candidate obstacle avoidance paths and operation control parameters. Furthermore, we first extract two boundary sections connecting the bottleneck region to the adjacent non-bottleneck passable region. We then calculate the feasible path distance from the robot's current position to the center of each boundary section. The boundary section with the smallest distance is defined as the bottleneck entrance, and the boundary section with the largest distance is defined as the bottleneck exit. Compared to simply using straight-line distance, using feasible path distance avoids incorrect entrance selection caused by wall obstruction. We then establish equidistant cross-sections along the longitudinal skeleton of the bottleneck region. On each cross-section, we obtain the midpoint between the left and right passable boundaries. We then perform curvature-constrained smoothing on the midpoint sequence to obtain the bottleneck centerline p. c (s), where s is the arc length along the centerline; To clarify the sign in the centerline formula, let: p c (s) = [x c (s), y c (s)]; Where p c(s) represents the ground coordinates of the point where the travel along the bottleneck centerline is an arc length of s, x c (s) and y c (s) represent the ground x-coordinate and ground y-coordinate of the coordinate point, respectively. s starts from 0 at the bottleneck entrance and is in meters. Then, offset paths are generated on both sides of the bottleneck centerline at preset offset intervals. The offset paths can be represented as: p d (s) = p c (s) + d×n(s); Where, p d (s) represents the ground coordinates of the offset path at the arc length parameter s, p c (s) represents the coordinates of the bottleneck centerline at the same arc length position, n(s) is the unit normal vector pointing laterally to one side of the centerline at point s, the length of the unit normal vector is 1 and it has no unit, d is the lateral offset distance with positive and negative directions, in meters. When d is positive, it indicates an offset to the side pointed to by n(s), when d is negative, it indicates an offset to the opposite side, and when d is 0, it is the bottleneck centerline. First, multiply the lateral offset d by the unit normal vector n(s), and then add it to the centerline coordinates p. c On (s), obtain the offset path coordinates p of the same arc length position. d (s) Place the robot's projection envelope point by point along the offset path, ensuring that it does not intersect with any obstacle at any position, that the boundary safety distance is not less than a preset value, and that the path curvature radius is not less than R. min The offset path is only retained when the offset amount is determined by the remaining width of the bottleneck cross section. The offset interval is an integer multiple of the map resolution. Therefore, the number of offset paths is determined by the actual channel width. Starting from the robot's current pose coordinates and heading angle, using a straight line segment with a radius not less than R... min The arc segment generates an approach path connecting the bottleneck entrance, making the position and tangential heading of the end of the approach path continuous with the bottleneck centerline or the starting point of the corresponding offset path. The approach path is spliced ​​with the centerline and each qualified offset path to obtain multiple candidate obstacle avoidance paths. For scenarios where it is not possible to turn around after entering the bottleneck and the cleaning task requires returning to the original entrance, each candidate obstacle avoidance path also records the return path segment and the path segment sequence from the cleaning endpoint on the bottleneck exit side to the bottleneck entrance in reverse along the corresponding path. Based on the ground material boundaries, path connection points, and driving direction switching points, the k-th candidate obstacle avoidance path is divided into multiple path segments. The operation control parameters of the i-th path segment are denoted as: U i =[v i Q i P s,i P c,i ]; Among them, U iLet v represent the vector of operation control parameters for the i-th path segment. i Q represents the travel speed of the i-th path segment, in meters per second. i P represents the output flow rate of the cleaning fluid for the i-th path segment, expressed in milliliters per second. s,i P represents the water absorption power of the i-th path segment, in watts. c,i The cleaning pressure for the i-th path segment is expressed in Newtons (N), where i represents the path segment number. The allowable range of each parameter is derived from the cleaning robot product parameters and actuator calibration results. The initial configuration value is read from the operation parameter table corresponding to the ground material, and then several discrete combinations are formed within the allowable range for evaluation of each candidate obstacle avoidance path. The cleaning fluid output is obtained after the flow meter calibrates the pump control command. The water absorption power is obtained by the water absorption motor control command combined with voltage and current feedback. The cleaning pressure is obtained by the load sensor on the lifting mechanism of the cleaning component. The grounding profiles B of the left and right drive wheels were obtained from the structural diagram of the cleaning robot and grounding imprint tests. w B w The ground contour B of the cleaning component is represented by the union of the ground contours of the left and right drive wheels in the robot's own coordinate system. This ground contour is obtained from the installation dimensions of the cleaning component and the static imprint test. c Let T be the rigid body transformation of the point at arc length s on the k-th candidate obstacle avoidance path, which transforms the robot's own coordinate system point to the ground coordinate system. pose,k (s), this transformation simultaneously includes the ground position of the robot reference point, the robot's yaw rotation, and the mounting offsets of the drive wheels and cleaning components relative to the reference point. To clarify this transformation, let any point in the robot coordinate system be z, and the ground coordinates of the reference point at arc length s along path k be: r k (s) = [x k (s), y k (s)], heading angle is φ k (s); T pose,k (s)[z]=r k (s) + Rot(φ) k (s))×z; Where z represents the coordinates of a two-dimensional point with the robot reference point as the origin, and r k (s) represents the ground position of the robot's reference point at arc length s along path k, x k (s) and y k (s) represent the x and y coordinates of the ground at that location, respectively, and φ k (s) represents the robot's heading angle relative to the ground coordinate system at that position, Rot(φ) k(s) is a two-dimensional rotation matrix that rotates the robot coordinate system orientation to the ground coordinate system orientation. The matrix form is the same as Rot(θ). It rotates the robot coordinate system point z along the current heading and then translates it to the reference point r. k (s), thus obtaining the position of this point in the ground coordinate system, the wheel track area and the cleaning fluid deposition area are respectively: W k =∪(0≤s≤L k )T pose,k (s) [B w ]; D k =∪(0≤s≤L k )T pose,k (s) [B c ]; Among them, W k D represents the entire wheel track region generated by the k-th candidate obstacle avoidance path. k L represents the total cleaning fluid deposition area generated by the k-th candidate obstacle avoidance path. k The total arc length of the k-th candidate obstacle avoidance path is represented by ∪ (0 ≤ s ≤ L). k ) indicates the path along the candidate obstacle avoidance path from s=0 to s=L k Take the union of all arc length positions with respect to the transformed grounding profile, T pose,k (s) [B w ] and T pose,k (s) [B c [ ] represents the regions after transforming the drive wheel grounding profile and the cleaning component grounding profile from the robot coordinate system to the ground coordinate system, respectively. Calculate D. k At the same time, it is also based on the Q of the path segment to which the arc length position belongs. i Whether the value is greater than zero records the sub-regions where liquid supply actually occurred, avoiding mistaking path segments that only absorb water but do not supply liquid as newly generated cleaning fluid deposition areas, due to B w and B c The mounting bias on the robot body is already included in T. pose,k In the transformation of (s), it is thus possible to determine whether the same ground location is first passed by the drive wheel or first covered by the cleaning component; Let the length of the i-th path segment be L. i The speed is v i If the start time of the first segment is t0, then the pure travel reference time of the starting point of the i-th path segment is: t move,i =t0+Σ (j=1) (i-1) (L) j / v j ); Among them, L jv is the arc length of the j-th completed path segment, in meters. j Σ represents the travel speed of the j-th completed path segment, in meters per second; t0 is the reference time of the starting point of the first path segment, in seconds; j is the path segment number used for accumulation; Σ (j=1) (i-1) This represents the sum of the travel times from segment 1 to segment (i-1). When i=1, this summation term is empty and takes the value 0. Therefore, t move,i This represents the pure travel reference time when the cleaning robot reaches the starting point of the i-th segment, ignoring parameter switching waiting time; The pure travel reference time at the location s from the starting point within this path segment is: t move,i (s) = t move,i +s / v i ; Where s is the arc length from the current position to the starting point of the i-th path segment, satisfying 0 ≤ s ≤ L i v i Let s / v be the speed of travel on the i-th path segment. i t represents the time required to travel from the starting point of this path segment to the current position. move,i (s) represents the pure travel reference time when the position is reached, ignoring parameter switching waiting time; If the switching of operation control parameters between the j-th path segment and the (j+1)-th path segment requires the robot to stop or slow down and wait for τ... j Then the actual arrival time of the corresponding position within the i-th path segment is: t a,i (s) = t move,i (s)+Σ (j=1) (i-1) τ j ; Where, τ j Σ represents the waiting time for switching job control parameters between the j-th path segment and the (j+1)-th path segment. (j=1) (i-1) τ j This represents the total switching wait time that has occurred before reaching the i-th segment, and is 0 when i=1. a,i (s) represents the actual arrival time including the switching wait. If the parameter can complete the switching during the journey without increasing the waiting time, then the corresponding τ j Take zero; Path arrival time in t move,i (s) serves as the baseline time series obtained from the path segment length and travel speed, while τ is recorded. j It also outputs the actual arrival time t used for wetting attenuation calculation. a,i(s) The path arrival time is calculated for the cleaning robot reference point, drive wheel grounding point, and cleaning component grounding area, respectively. The time difference is determined by the installation offset of each component and the travel speed. Subsequent calculations of travel time directly sum the recorded pure travel time and τ. j There is no need to set a separate switching time for different calibers, thus avoiding omissions or duplicate timing; For example, using a 2.40-meter bottleneck area, the bottleneck width is reduced by 0.60 meters to 0.26 meters. Zero offset and ±0.09-meter offset are set to obtain candidate obstacle avoidance paths A, B, and C. The lengths of the approach segment, the forward segment of the bottleneck, and the return segment of candidate obstacle avoidance path A are 1.60 meters, 2.40 meters, and 2.40 meters, respectively. The lengths of candidate obstacle avoidance path B are 1.64 meters, 2.43 meters, and 2.43 meters, respectively, due to the two smooth connections between the offset path and the approach path. The lengths of candidate obstacle avoidance path C are 1.62 meters, 2.41 meters, and 2.41 meters, respectively, due to the two smooth connections between the offset path in another direction and the approach path. The lengths are obtained by the cumulative arc length of the straight and circular segments of the path. The entry section of path A is an epoxy floor, configured with a travel speed of 0.40 m / s, a cleaning fluid output of 10 ml / s, a water absorption power of 420 watts, and a cleaning pressure of 180 N. The bottleneck forward ceramic tile section is configured with 0.25 m / s, 9 ml / s, 500 watts, and 160 N. The ceramic tile return section is configured with 0.20 m / s, 8 ml / s, 520 watts, and 140 N. Path B employs a conservative parameter combination of lower speed and lower liquid supply. The inlet section is configured with 0.38 m / s, 8 ml / s, 450 W and 170 N, respectively; the bottleneck forward section is configured with 0.23 m / s, 7 ml / s, 530 W and 150 N, respectively; and the return section is configured with 0.18 m / s, 6 ml / s, 550 W and 130 N, respectively. Path C adopts a more aggressive combination of short-duration parameters. The approach section is configured with 0.45 m / s, 12 ml / s, 380 W and 200 N in sequence. The bottleneck forward section is configured with 0.40 m / s, 14 ml / s, 350 W and 220 N in sequence. The return section is configured with 0.45 m / s, 14 ml / s, 350 W and 220 N in sequence. Each set of values ​​is given in the order of travel speed, cleaning fluid output, water absorption power and cleaning pressure, and all are within the calibration allowable range of the prototype actuator. Taking the start time of the first segment of path A as time zero, the pure travel reference time for the cleaning robot to reach the bottleneck entrance is: 1.60 / 0.40 = 4.00 seconds; The pure travel reference time for reaching the clean endpoint on the bottleneck exit side is: 4.00 + 2.40 / 0.25 = 13.60 seconds; The pure travel reference time for returning to the bottleneck entry is: 13.60 + 2.40 / 0.20 = 25.60 seconds; The prototype calibration table records that the transition from the entry section to the bottleneck forward section and from the bottleneck forward section to the return section both require a 0.80-second pause. Therefore, the actual time to reach the cleaning endpoint at the bottleneck exit side is 14.40 seconds, and the actual time to return to the bottleneck inlet is 27.20 seconds. The ground imprint test measured the effective ground width of the cleaning component to be 0.46 meters, and the ground width of a single drive wheel to be 0.08 meters. By scanning the installation position within the coordinate system of the cleaning robot, the corresponding wheel track area and cleaning fluid deposition area are obtained. The geometric area, parameter sequence, pure travel reference time, parameter switching time, and actual arrival time together constitute the data input for calculating the time-series wettability.

[0020] Please see Figure 1 and Figure 4 The present invention provides an embodiment of an obstacle avoidance path optimization method for cleaning robots in multi-obstacle scenarios, which calculates the temporal wetness of the cleaning fluid deposition area by means of path arrival time, ground material and operation control parameters, and obtains the lower bound of the friction coefficient of the wheel track area from the pre-established wetness-friction coefficient mapping relationship by means of the temporal wetness. Furthermore, for the i-th path segment, let the output volume of the cleaning fluid be Q. i The speed of travel is v i If the effective grounding width of the cleaning component is b, then the liquid supply per unit area is q. in for: q in =Q i / (v i ×b); Where, q in Q is the liquid supply volume per unit area, expressed in milliliters per square meter. i v represents the output flow rate of the cleaning fluid for the i-th path segment, expressed in milliliters per second. i The travel speed of this path segment is in meters per second, b is the effective grounding width of the cleaning component in meters, and v is the travel speed of this path segment. i ×b represents the area covered by the cleaning component per unit time; therefore, Q i / (v i ×b) This allows the flow rate to be converted into the liquid supply per unit area. The recovery efficiency of the water absorption component on different floor materials is pre-calibrated through quantitative liquid supply tests. Let the floor material type be m and the water absorption power be P. s The coefficient of recovered liquid volume corresponding to the unit water absorption energy is k. r,m Then the amount of liquid recovered per unit area q out for: q out =k r,m ×P s / (vi ×b); Where, q out k represents the amount of liquid recovered per unit area. r,m P is the unit water absorption energy recovery coefficient for ground material m. s For water absorption power, k r,m ×P s This represents the amount of liquid recovered per unit time, divided by the coverage area v per unit time. i ×b yields the amount of recovered liquid per unit area, and the coefficient k r,m The method for obtaining the residual liquid mass q is as follows: a standard sample of material m is passed through a known volume of liquid at multiple water absorption powers and a fixed speed. The mass of residual liquid before and after passage is weighed, and the recovered volume is divided by the water absorption energy to obtain the residual liquid mass q per unit area. r for: q r =max(0, q) in -q out ); Where, q r The residual liquid volume per unit area, max(0, q) in -q out This indicates that the maximum value is taken between 0 and the difference between supply and recovery, so that the residual liquid amount will not be negative. When the recovery amount is greater than the supply amount, q r Taking 0 as the threshold, static absorption tests were conducted for different ground materials and cleaning pressures. A known amount of residual liquid per unit area was applied to the sample, and the proportion of wetted surface area or the mass of liquid on the surface was measured after a preset waiting time. The liquid absorption conversion coefficient α was then obtained through fitting. m (P) c The initial wettability H0 formed after each cleaning fluid deposition is: H0=a m (P) c )×q r ; Where H0 represents the initial wettability formed by the initial deposition of the cleaning fluid, and wettability is a normalized dimensionless index, a m (P) c ) represents the ground material m under cleaning pressure P c Liquid absorption conversion coefficient, P c For cleaning pressure, for the same location x, the j-th deposition time t is obtained based on the arrival time sequence of the cleaning component path. j The wetting attenuation parameter λ was obtained based on the material drying test. m Then the time-series wetness at time t is: H(x,t) = Σ{H 0,j (x)×e -λm×(t-tj)}, t≥t j ; Where H(x, t) represents the temporal humidity at ground location x at time t, x is the ground coordinate location, t is the time at which the humidity is queried, and Σ represents the time sequence of all conditions satisfying t. j Sum of cleaning fluid deposition events ≤ t and covering position x, H 0,j (x) represents the initial wettability at location x where the j-th deposition occurs, t j Let λ be the actual time when the j-th deposition reaches position x. m e is the wetting attenuation parameter corresponding to ground material m. -λm×(t-tj) This indicates that the j-th deposition occurred from time t. j The decay rate up to time t, tt j Let x be the elapsed time. If the same location is visited repeatedly, then all depositional events occurring before the current time are attenuated separately and then summed. If location x is outside the coverage area of ​​a certain depositional event, then H is... 0,j (x) is set to 0. If the sum exceeds the maximum wettability sampling state of the calibration mapping relationship, the lower bound of the friction coefficient corresponding to the maximum sampling state is adopted, and the data outside the range of the mapping relationship is not extrapolated. When establishing the mapping relationship between wettability and friction coefficient, samples identical to those used on-site were prepared for each type of flooring material, such as ceramic tiles and epoxy flooring. Multiple wettability sampling states were set up, ranging from a dry state to the maximum permissible wettability state. Wettability was obtained through quantitative liquid supply and weighing verification. A robot's drive wheel of the same model was pressed onto the sample, and a constant normal load N0 was maintained using a load sensor. N0 represents the normal load of the drive wheel on the sample. The drive wheel torque was gradually increased, and the slip ratio was calculated using the encoder wheel circumference speed and the ground speed from an external displacement sensor. When the slip ratio reached 5% and remained so for 0.3 seconds, the slip threshold was determined to have been reached, and the critical output torque T was recorded. c T c This represents the output torque of the drive wheel when the slip threshold is just reached. Let the effective radius of the drive wheel be r. w r w Let F represent the effective distance from the center of the drive wheel axle to the contact surface, then the critical traction force is F. c The measured friction coefficient μ and the actual friction coefficient are respectively: F c =T c / r w ; μ=F c / N0; Among them, F c For the critical traction force, T c For the critical output torque, r w The effective radius of the drive wheel is given by , μ is the measured friction coefficient obtained experimentally under this wettability condition, N0 is the constant normal load, and F... cThis represents the conversion of the drive wheel torque into the tangential traction force at the contact surface. μ represents the ratio of the critical traction force to the normal load. Each wettability sampling condition is repeated at least five times. The resulting wettability and friction coefficient constitute the calibration sample. Monotonic fitting is performed according to the constraint that the friction coefficient does not increase with increasing wettability to obtain the fitted curve μ. f (H), where H is the humidity, μ f (H) is the average friction coefficient given by the fitted curve at the humidity H. The standard error σ(H) for each humidity interval is calculated by repeated experiments and fitting residuals. σ(H) represents the statistical measure of the fitting uncertainty at the humidity H. For example, when using a one-sided 95% confidence level, the lower bound is: μ l (H) = max[0, μ] f [(H)-1.645×σ(H)]; Where, μ l (H) represents the lower bound of the friction coefficient corresponding to the wettability H, μ f (H) is the monotonic fitted friction coefficient at wetness H, σ(H) is the fitting standard error at that wetness, and 1.645 is the normal distribution coefficient corresponding to the 95% confidence level on one side. The lower boundary is used as the wetness-friction coefficient mapping relationship for material m. Using the lower boundary instead of the fitted mean can preserve a safety margin for sample differences, tire wear, and measurement errors. When evaluating candidate obstacle avoidance paths, the time of arrival of the driving wheel is considered within the wheel track area. w Query H(x,t) w ), where t w Given the actual arrival time of the drive wheel contact position x, obtain μ for each wheel track position. l For a given path segment, the minimum value among all wheel track positions in that path segment is used as the lower bound of the friction coefficient. For example, the unit water absorption energy recovery liquid volume coefficient k for ceramic tiles r,m The wettability conversion factor is 0.010 ml / J. a m (P) c = 0.012 × (1 + 0.0015 × P) c ); Wetting attenuation parameter λ m The value is 0.035 seconds. The parameters are from quantitative recovery tests, absorption tests under different pressures, and natural decay weighing tests. 0.012 is the basic absorption conversion coefficient of the ceramic tile under zero cleaning pressure, and 0.0015 is the relative correction coefficient for each 1 Newton increase in cleaning pressure. For the bottleneck forward segment of path A, Q i 9 ml / s, v i=0.25 m / s, b = 0.46 m, P s For 500 watts, P c Given 160 Newtons, the calculation yields: q in =9 / (0.25×0.46) = 78.26 ml / m²; q out =0.010×500 / (0.25×0.46)=43.48 ml / m²; q r =78.26-43.48=34.78 ml / m²; H 0,1 =0.012×(1+0.0015×160)×34.78=0.518; For the return segment, Q i 8 ml / s, v i =0.20 m / s, b = 0.46 m, P s For 520 watts and P c Given 140 Newtons, q is calculated. in It is 86.96 ml / m², q out It is 56.52 ml / m², q r It is 30.44 ml / m², H 0,2 The value is 0.442. In the prototype, the center of the cleaning component is located 0.30 meters behind the drive wheel axis, at the midpoint of the bottleneck. After accounting for the switching time of two 0.80-second parameters, the cleaning component forms the first deposit in 10.80 seconds when moving forward. When returning in the opposite direction, the cleaning component arrives before the drive wheel relative to the direction of movement, forming the second deposit in 19.70 seconds. The drive wheel arrives in 21.20 seconds. The time is obtained by dividing the actual arrival time and the 0.30-meter installation offset by the corresponding speed. Therefore, the timing wetting degree when the drive wheel reaches this position on the return is: H = 0.518 × e -0.035×(21.20-10.80) +0.442×e -0.035×(21.20-19.70) =0.779; After monotonic fitting and subtracting the one-sided 95% confidence interval, the ceramic tile calibration samples obtained μ within the calibration range of this embodiment. l (H) = max[0.180, 0.620 - 0.380 × H], where H is the maximum humidity sampling state. max It is 1.16, where H max This indicates the maximum wettability that the calibration test can cover. Substituting H as 0.779, we get the lower limit of the friction coefficient when returning from this position as 0.324. When traveling in the forward direction, the drive wheel reaches this position before the cleaning component, and the current liquid supply deposition has not yet occurred. Therefore, the calculation should be based on the existing deposition event at that time, and the wettability that will occur later should not be used in advance. As another illustration, the bottleneck forward segment of path C is calculated using parameters of 14 ml / s, 0.40 m / s, 350 watts, and 220 N, yielding q. r 57.07 ml / m², H 0,1 The value is 0.911. The return segment is calculated with parameters of 14 ml / s, 0.45 m / s, 350 watts, and 220 N, resulting in q. r 50.72 ml / m², H 0,2 The value is 0.810. Based on the two parameter switching times of 0.40 seconds for path C and the same installation offset of 0.30 meters, when the return wheel track arrives at the midpoint of the bottleneck, the sum of the attenuation results of the two depositions yields a temporal wetness of 1.55, which exceeds H. max Therefore, by querying the mapping relationship according to the maximum humidity sampling state 1.16, the lower bound of the friction coefficient is obtained as 0.180. This shows that the 0.324 of path A and the 0.180 of path C are both calculated by their respective operation control parameters, actual arrival time sequence and the calibration relationship of the same tile. Clearly distinguishing between the pure travel reference time sequence and the actual arrival time sequence can avoid missing the humidity decay that occurs during the parameter switching waiting period.

[0021] Please see Figure 1 and Figure 4 The present invention provides an embodiment of a cleaning robot obstacle avoidance path optimization method for multi-obstacle scenarios. Based on the lower bound of the friction coefficient, the driving force margin and braking force margin corresponding to the cleaning robot entering the bottleneck area and returning to the entrance of the bottleneck area along the candidate obstacle avoidance path are calculated respectively. Candidate obstacle avoidance paths whose mechanical margins do not meet the motion constraints are screened out. The target obstacle avoidance path with the shortest passage time is selected from the remaining candidate obstacle avoidance paths. The cleaning robot is controlled to perform cleaning operations according to the target obstacle avoidance path and the corresponding operation control parameters. Furthermore, calculation positions are established along each candidate obstacle avoidance path at fixed arc length intervals, as well as at path segment endpoints, speed switching points, and ground material switching points. For each calculation position, the lower bound μ of the friction coefficient at the moment the drive wheel arrives at that position is calculated. l And obtain the normal load N of the drive wheel. d μ l It is the wettability mapping result corresponding to the calculated location and the arrival time of the drive wheel, not a fixed constant for the entire path, N. d N represents the effective normal load of the driving wheel on the ground. d The static axle load is corrected by measuring the load cell after the robot is loaded with the rated cleaning fluid. During operation, it can also be conservatively calculated using the minimum possible drive axle load based on the acceleration given by the inertial measurement unit and the pressure given by the cleaning component load sensor. The upper limit of the adhesion force F can be used. a for: Fa =μ l ×N d ; Among them, F a μ represents the maximum tangential force that can be provided by wheel-to-ground adhesion. l The lower bound of the dimensionless coefficient of friction, N d For the effective normal load, let the speed commands for the (i-1)th path segment and the ith path segment be v, respectively. i-1 and v i The speed change time required to complete the switch is Δt. i Then the acceleration a along the current direction of travel is: a = (v i -v i-1 ) / Δt i ; Where a is the longitudinal acceleration, v i and v i-1 Let Δt be the speed of travel between adjacent path segments. i This represents the actual transition time from the previous speed to the current speed. When the speed decreases, 'a' is negative. Let the mass of the cleaning robot be M, and the rolling resistance coefficient of the ground material be c. r c r The coefficients are dimensionless, the acceleration due to gravity is g, and the cleaning pressure is P. c The calibration factor for converting cleaning pressure into horizontal cleaning resistance is k. c If the unit is dimensionless, then the demand driving force F d for: F roll =c r ×M×g; F clean =k c ×P c ; F d =M×|a|+F roll +F clean ; Among them, F roll F represents rolling resistance. clean The horizontal cleaning resistance between the cleaning components and the ground is represented by 'a', the longitudinal acceleration caused by adjacent speed commands is represented by '|a|', and the absolute value of the acceleration is represented by '|a|', with acceleration and deceleration conservatively factored in based on the required driving force. 'M' is the total mass of the cleaning robot, 'g' is the acceleration due to gravity, and 'c' is the acceleration due to gravity. r The value was obtained by conducting a low-speed traction test on a surface of the corresponding material after the cleaning component was turned off. c The driving force margin M was obtained by fitting the difference in steady-state traction force when the cleaning component is lifted and when it lands with different pressures. d for: Md =F a -F d ; Among them, M d M represents the remaining force after deducting the required driving force from the available adhesion force. d The larger the value, the greater the risk margin of drive wheel slippage. At locations where braking is required before a bottleneck or stopping point, the braking distance d along the path to the stopping point is determined based on the current speed v and the remaining braking distance d. r Calculate the required braking force F b And obtain the braking force margin M b : F b =M×v 2 / (2×d r ); M b =F a -F b ; Among them, F b This represents the theoretical braking force required to reduce the current speed to a stop within the remaining braking distance, where v represents the current speed at the calculated position, and d represents the braking force. r This represents the remaining distance along the current direction of travel to the stopping point or bottleneck boundary, in meters and must be greater than 0. M is the total mass of the cleaning robot. b This represents the braking force margin after deducting the required braking force from the available adhesion force. The braking calculation does not utilize the auxiliary effects of rolling resistance and cleaning resistance on deceleration, and is therefore a conservative calculation. For the process of the robot moving from the bottleneck entrance to the inside, the calculation is performed point by point in the order of the forward path segment. For the process of the robot returning from the inside of the bottleneck to the entrance, the timing wettability, upper limit of adhesion force, driving force margin and braking force margin are recalculated in the order of the reverse path segment. The minimum value of each margin is taken in the entry process and the return process to prevent the low adhesion position in the middle of the path from being missed by only checking the entrance or exit endpoint. The safety margin threshold M of the driving force was determined in advance through start-up, steady-state driving, and emergency braking tests of a fully loaded prototype. d,min and braking force safety margin threshold M b,min , of which M d,min M represents the minimum permissible driving force margin. b,min This represents the minimum allowable braking force margin. The threshold must at least cover the driving torque error, shaft load fluctuation, and friction calibration error. In this embodiment, M is taken as the upper bound of the error from multiple tests. d,min For 40 Newtons, M b,min For a value of 35 Newtons, the k-th candidate obstacle avoidance path is retained only if it simultaneously meets the following conditions: M d,in ≥M d,min ; M b,in≥M b,min ; M d,r ≥M d,min ; M b,r ≥M b,min ; Where in represents the robot's entry direction from the bottleneck entrance to the inside of the bottleneck, r represents the robot's return direction from the inside of the bottleneck back to the bottleneck entrance, and M d,in M b,in M d,r and M b,r These represent the minimum driving force margin and minimum braking force margin in the corresponding direction, and in all calculated positions, respectively. If any condition is not met, the candidate obstacle avoidance path is eliminated. The margins in each direction are the minimum values ​​in all calculated positions in the corresponding direction. For example, the cleaning robot has a mass M of 120 kg and a conservative normal load N on the drive wheel. d The rolling resistance coefficient of the ceramic tile is 780 Newtons. r The cleaning resistance calibration coefficient k is 0.018. c For the bottleneck midpoint in the return process of path A, μ is 0.35. l Given a velocity of 0.324, a speed of 0.20 m / s, an absolute value of acceleration corresponding to the velocity change of 0.05 m / s², a cleaning pressure of 140 N, and a remaining braking distance of 0.12 m, then: F a =0.324 × 780 = 252.72 Newtons; F roll =0.018×120×9.81=21.19 Newtons; F clean =0.35×140=49.00 Newtons; F d =120×0.05+21.19+49.00=76.19 Newtons; M d =252.72-76.19=176.53 Newtons; F b =120×0.20 2 / (2 × 0.12) = 20.00 Newtons; M b =252.72 - 20.00 = 232.72 Newtons; Therefore, both margins at this calculation location meet the threshold, and the high-speed, high-liquid-supply combination of path C returns to the calculation location at a point using the μ calculated in the example. lThe acceleration is 0.180, the speed is 0.45 m / s, the absolute value of the acceleration is 0.18 m / s², the cleaning pressure is 220 N, the remaining braking distance is 0.10 m, and F... a 140.40 Newtons, F d It is 119.79 Newtons, M d For 20.61 Newtons, F b 121.50 Newtons, M b The value is 18.90 N. The driving force margin of path C is less than 40 N and the braking force margin is less than 35 N, so it is screened out. After calculating the full path for candidate obstacle avoidance paths A, B, and C at 0.05-meter arc length intervals, paths A and B are retained because their mechanical margins in both the entry and return directions are not less than the corresponding thresholds. Path C is eliminated because its calculated position does not meet the motion constraints. Subsequently, the travel time for each retained path is calculated, assuming the length of the i-th path segment is L. i The speed is v i The actual switching time of the operation control parameters between the adjacent i-th path segment and the (i+1)-th path segment is τ. i If the total number of path segments is n, then the travel time T of candidate obstacle avoidance path k is... k for: T k =Σ (i=1) n (L) i / v i )+Σ (i=1) (n-1) τ i ; Among them, T k This represents the total travel time of the k-th candidate obstacle avoidance path from its entry point to its return to the bottleneck entrance, where k is the candidate obstacle avoidance path number, n is the total number of path segments contained in the candidate obstacle avoidance path, and L... i and v i Let τ be the arc length and travel speed of the i-th path segment, respectively. i Let τ be the parameter switching waiting time between segment i and segment i+1. The first term is the sum of the pure travel time of all path segments, and the second term is the sum of the switching waiting times of all adjacent path segments. If the switching process does not increase the waiting time, then the corresponding τ i Set to 0; The pure travel time of path A is: 1.60 / 0.40 + 2.40 / 0.25 + 2.40 / 0.20 = 25.60 seconds; The switching points for both parameters require 0.80 seconds each, as determined by the actuator; therefore, the passage time is 27.20 seconds. The pure travel time of path B is: 1.64 / 0.38 + 2.43 / 0.23 + 2.43 / 0.18 = 28.38 seconds; The two switching points take a total of 1.20 seconds, so the passage time is 29.58 seconds. The passage time of path C calculated according to the speed configuration is 15.78 seconds, but it is screened out because the mechanical margin does not meet the motion constraints and is not included in the shortest time comparison. Therefore, path A with a passage time of 27.20 seconds is selected from the remaining candidate obstacle avoidance paths as the target obstacle avoidance path. The operation control parameters corresponding to each path segment of path A are determined as target operation control parameters and stored in the robot controller according to the connection order of the path segments. When performing cleaning operations, the controller projects the real-time pose onto the target obstacle avoidance path. Based on the projection arc length, path segment boundary and current travel direction, the robot's path segment is determined. The corresponding travel speed, cleaning fluid output, water absorption power and cleaning pressure are read from the target operation control parameters. The speed is controlled by the drive wheel in a closed loop, the cleaning fluid output is controlled by the pump and flow meter in a closed loop, the water absorption power is controlled by the water absorption motor power feedback, and the cleaning pressure is controlled by the load sensor and lifting actuator in a closed loop. The cleaning robot enters the bottleneck area according to the target obstacle avoidance path and its parameter sequence, completes cleaning and returns to the bottleneck entrance. At the same time, it obtains a shorter passage time in the candidate scheme that meets the motion safety constraints.

[0022] Working principle: First, the outlines of obstacles, ground materials, and pose of the cleaning robot in the working area are collected. Then, the minimum turning radius of the cleaning robot is scanned to form the minimum turning envelope area. The passable area is determined according to the obstacle outline. The continuous area that is jointly defined by multiple obstacles and cannot accommodate the minimum turning envelope area is identified as the bottleneck area. Then, candidate obstacle avoidance paths are generated from the cleaning robot pose to pass through the bottleneck area. The operating control parameters such as travel speed, cleaning fluid output, water absorption power, and cleaning pressure are configured for each candidate obstacle avoidance path. The arrival time of the drive wheel track area, the cleaning fluid deposition area, and the position of each path are further determined. Based on the amount of cleaning fluid deposited, the amount of water absorbed and recovered, the cleaning pressure, the ground material, and the duration of deposition, the temporal wetness of the wheel track area when the robot arrives is calculated. The lower bound of the friction coefficient is obtained through a pre-established wetness-friction coefficient mapping relationship, which is calibrated through drive wheel slippage tests under different wetness conditions. The upper bound of the adhesion force is calculated based on the lower bound of the friction coefficient and the normal load of the drive wheel. Then, combined with speed changes, rolling resistance, cleaning resistance, and remaining braking distance, the driving force margin and braking force margin when the robot enters the bottleneck area and returns to the bottleneck entrance are calculated respectively. Candidate obstacle avoidance paths that do not meet the motion constraints are screened out. Finally, the travel time of the retained path segments and the switching time of the operation control parameters are compared, and the target obstacle avoidance path with the shortest passage time is selected. The cleaning robot is controlled to complete the cleaning operation according to the operation control parameters corresponding to each path segment.

[0023] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention.

Claims

1. A method for optimizing obstacle avoidance paths for cleaning robots in multi-obstacle scenarios, characterized in that, include: S1. Collect obstacle contours, ground material and cleaning robot pose. Based on the minimum turning radius, scan the cleaning robot to obtain the minimum turning envelope area. Determine the passable area according to the obstacle contour. Identify the passable area that is jointly limited by multiple obstacles and cannot accommodate the minimum turning envelope area as the bottleneck area. S2. Generate candidate obstacle avoidance paths through the bottleneck area based on the robot's pose. Configure operation control parameters consisting of travel speed, cleaning fluid output, water absorption power, and cleaning pressure for each candidate obstacle avoidance path. Determine the wheel track area, cleaning fluid deposition area, and path arrival sequence based on the candidate obstacle avoidance path and operation control parameters. S3. Calculate the temporal wetness of the cleaning fluid deposition area by means of path arrival time, ground material and operation control parameters, and obtain the lower bound of the friction coefficient of the wheel track area from the pre-established wetness-friction coefficient mapping relationship by means of temporal wetness. S4. Based on the lower bound of the friction coefficient, calculate the driving force margin and braking force margin corresponding to the cleaning robot entering the bottleneck area and returning to the entrance of the bottleneck area along the candidate obstacle avoidance path, and screen out any candidate obstacle avoidance path whose mechanical margin does not meet the motion constraints. S5. Select the target obstacle avoidance path with the shortest travel time from the retained candidate obstacle avoidance paths, and control the cleaning robot to perform cleaning operations according to the target obstacle avoidance path and the corresponding operation control parameters.

2. The multi-obstacle scene oriented cleaning robot obstacle avoidance path optimization method according to claim 1, characterized in that: The specific process for obtaining the minimum turning envelope region is as follows: Obtain the initial projection envelope of the cleaning robot in the ground coordinate system. Establish a turning trajectory from the initial heading to the reverse heading based on the minimum turning radius. Transform the center coordinates and heading angle of the initial projection envelope along the turning trajectory. Take the union of all transformed projection envelopes to obtain the minimum turning envelope region.

3. The multi-obstacle scene oriented cleaning robot obstacle avoidance path optimization method according to claim 2, characterized in that: The specific process for identifying the bottleneck region is as follows: The obstacle contours are mapped onto the ground coordinate system. The area enclosed by the obstacle contours is removed from the working area to obtain the passable area. The projection envelope of the cleaning robot is moved within the passable area. The positions where the projection envelope does not intersect with the obstacle contours form a set of straight-line passable positions. The minimum turning envelope area is moved within the set of straight-line passable positions. The positions where the minimum turning envelope area intersects with the obstacle contours form a set of turning-restricted positions. The continuous position area in the set of turning-restricted positions, which is jointly defined by multiple obstacles, is determined as the bottleneck area.

4. The multi-obstacle scene oriented cleaning robot obstacle avoidance path optimization method according to claim 3, characterized in that: The specific process for generating candidate obstacle avoidance paths that pass through the bottleneck region is as follows: Extract the boundary sections at both ends of the bottleneck region, determine the center of the boundary section of the adjacent cleaning robot pose as the bottleneck inlet, and determine the center of the other boundary section as the bottleneck outlet. Generate the bottleneck centerline based on the midpoint of each cross section between the bottleneck inlet and the bottleneck outlet. Generate offset paths that satisfy the projection envelope passage constraints on both sides of the bottleneck centerline. Generate an approach path connecting the bottleneck inlet based on the position coordinates and heading angle of the cleaning robot pose. Concatenate the approach path with the bottleneck centerline and the offset path respectively to obtain the candidate obstacle avoidance path.

5. The multi-obstacle scene oriented cleaning robot obstacle avoidance path optimization method according to claim 1, characterized in that: The specific process for determining the wheel track region, the cleaning fluid deposition region, and the path arrival time sequence is as follows: The cleaning robot includes drive wheels and cleaning components. The candidate obstacle avoidance path is divided into multiple path segments according to the ground material. The travel speed, cleaning fluid output, water absorption power and cleaning pressure are configured for each path segment to form a sequence of operation control parameters. The ground contour of the drive wheels is scanned along each path segment to obtain the wheel track area. The ground contour of the cleaning components is scanned along each path segment to obtain the cleaning fluid deposition area. The travel time of each path segment is calculated based on the length and travel speed of each path segment. The travel time of each path segment is accumulated according to the connection order of the path segments to obtain the path arrival time sequence.

6. The obstacle avoidance path optimization method for cleaning robots in multi-obstacle scenarios according to claim 5, characterized in that: The specific process for calculating the temporal wettability of the cleaning fluid deposition area is as follows: The liquid supply per unit area is calculated based on the cleaning fluid output, travel speed, and grounding width of the cleaning components. The liquid recovery per unit area is determined based on the water absorption power. The difference between the liquid supply per unit area and the liquid recovery per unit area is determined as the residual liquid per unit area. The residual liquid per unit area is converted into initial wettability based on the cleaning pressure and the liquid absorption parameters corresponding to the ground material. The duration after each cleaning fluid deposition is determined based on the path arrival time sequence. The initial wettability is attenuated using the wettability attenuation parameters corresponding to the ground material. The attenuation results of the repeated traversal locations are added together to obtain the time-series wettability.

7. The obstacle avoidance path optimization method for cleaning robots in multi-obstacle scenarios according to claim 1, characterized in that: The specific process of pre-establishing the mapping relationship between wettability and friction coefficient is as follows: Multiple humidity sampling states are set for each type of ground material. Under constant normal load, the output torque of the drive wheel is increased step by step. The critical traction force when the drive wheel reaches the slip threshold is recorded. The friction coefficient is obtained by calculating the ratio of the critical traction force to the constant normal load. The humidity sampling states and the friction coefficient are combined to form a calibration sample. The calibration sample is monotonically fitted according to the direction of increasing humidity. The friction coefficient confidence interval is generated based on the fitting error. The lower boundary of the friction coefficient confidence interval is determined as the humidity-friction coefficient mapping relationship corresponding to the ground material.

8. The obstacle avoidance path optimization method for cleaning robots in multi-obstacle scenarios according to claim 5, characterized in that: The specific process for calculating the driving force margin and braking force margin is as follows: The upper limit of adhesion force is calculated based on the lower limit of the friction coefficient and the normal load of the drive wheel. The required driving force is calculated based on the changes in travel speed, rolling resistance, and cleaning resistance generated by cleaning pressure. The driving force margin is obtained by subtracting the required driving force from the upper limit of adhesion force. The required braking force is calculated based on the travel speed and the remaining braking distance in the bottleneck area. The braking force margin is obtained by subtracting the required braking force from the upper limit of adhesion force. The driving force margin and braking force margin corresponding to entering the bottleneck area and returning to the bottleneck area entrance are calculated separately according to the forward path segment order and the reverse path segment order of the candidate obstacle avoidance path. 9.The multi-obstacle scene oriented cleaning robot obstacle avoidance path optimization method according to claim 8, characterized in that: The process for selecting the target obstacle avoidance path is as follows: The driving force margin and braking force margin are compared with their respective preset safety margin thresholds. Candidate obstacle avoidance paths with all mechanical margins not less than the safety margin threshold are retained. The travel time of each retained candidate obstacle avoidance path segment and the switching time of operation control parameters are calculated. The sum of the travel time of the path segment and the switching time of operation control parameters is determined as the passage time. The candidate obstacle avoidance path with the shortest passage time is selected as the target obstacle avoidance path.

10. The obstacle avoidance path optimization method for cleaning robots in multi-obstacle scenarios according to claim 9, characterized in that: Each segment of the target obstacle avoidance path corresponds to a corresponding operation control parameter. The corresponding operation control parameter is determined as the target operation control parameter. The current path segment is determined based on the robot's pose. The travel speed, cleaning fluid output, water absorption power, and cleaning pressure corresponding to the current path segment are read from the target operation control parameter, and the robot is controlled to perform cleaning operations.