A method and system for planning a full coverage path for a mowing robot

Abstract

The application relates to a full coverage path planning method and system for a mowing robot, which first generates a global reference path in the shape of a Chinese character 'Gong' (bowed). When an unknown obstacle is detected by a sensor during the operation of the robot along the global path, an improved DWA local planner is triggered, a coverage following factor is added to the evaluation function of the improved DWA local planner, the lateral error between the obstacle avoidance trajectory and the global path is minimized, and the smooth transition of obstacle avoidance and tracking is realized. Meanwhile, the system records the path segments that are not covered due to deviation for obstacle avoidance in real time, establishes a missed cleaning task queue, and automatically schedules the robot to perform secondary missing area cleaning in the queue after the global main path is traversed. The application effectively reduces the missed cleaning area and regression oscillation in the obstacle avoidance process, significantly improves the full coverage rate and smoothness of operation in a complex dynamic environment, reduces hardware loss, and is suitable for full coverage path planning of various mobile cleaning and mowing robots.

Description

Technical Field

[0001] This invention relates to the field of agricultural robots and intelligent operating equipment, specifically to a method and system for full-coverage path planning of a sweeping and mowing robot that integrates environmental memory and an improved DWA algorithm. Background Technology

[0002] When existing cleaning and mowing robots perform full-coverage operations in complex environments, they generally suffer from problems such as a disconnect between local obstacle avoidance and global coverage tasks, slow or oscillating path alignment after obstacle avoidance, and inability to compensate for missed areas due to a lack of environmental memory.

[0003] 1. Disconnect between local obstacle avoidance and global coverage tasks: In traditional robots (such as robotic vacuum cleaners and lawnmowers), global planning and local obstacle avoidance are usually two separate modules when performing full-coverage path planning. When the robot encounters an obstacle and triggers a local obstacle avoidance algorithm (such as traditional DWA, Dynamic Window Approach), the algorithm's priority is often set entirely to "safe obstacle avoidance" and "quickly reaching the target point." This causes the robot to arbitrarily deviate from the original "bow"-shaped global path during obstacle avoidance. Due to the lack of "centripetal force" constraint on the original path, the robot often detours too far, resulting in large unsweeped vacuum areas around the obstacle and reducing operational efficiency.

[0004] 2. Slow or oscillating path alignment after obstacle avoidance: Existing technologies typically rely on simple PID controllers or pure tracking algorithms to attempt to return to the global path after obstacle avoidance. Because the local obstacle avoidance trajectory and the global tracking trajectory are mathematically discontinuous, the robot often exhibits significant angular oscillations or "S-shaped" serpentine movements when returning to the original path, leading to a significantly increased rate of repetitive tasks during the return process. In complex environments such as lawns or indoors, this uneven transition also increases wear and tear on the hardware motors.

[0005] 3. Lack of environmental memory leads to insufficient coverage: Most robotic cleaning systems treat obstacle avoidance as a "temporary disturbance," assuming that the task continues once obstacle avoidance is complete. The system lacks a quantitative assessment of the "coverage cost" resulting from obstacle avoidance. In complex, dynamic scenarios with many obstacles, the robot leaves numerous scattered missed areas due to frequent obstacle avoidance. Because the system lacks "environmental memory" and a "secondary compensation mechanism" for these missed areas, the final coverage rate after the job is completed is far lower than the theoretical value, failing to achieve true 100% coverage. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a full-coverage path planning method and system for a sweeping and mowing robot. By integrating environmental memory and an improved DWA algorithm, it effectively reduces the missed sweeping area and regression oscillation during obstacle avoidance, and significantly improves the full coverage rate and operational smoothness in complex dynamic environments.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for full-coverage path planning of a sweeping and mowing robot is proposed. First, a "bow"-shaped global path is generated using a ox-plowing-style unit decomposition method. When the robot operates along the global path, if the sensor detects an unknown obstacle, an improved DWA local planner is triggered. The trajectory evaluation function adds a coverage following factor to minimize the lateral error between the obstacle avoidance trajectory and the global path, and outputs the optimal obstacle avoidance trajectory to achieve a smooth transition between obstacle avoidance and tracking. At the same time, the system records the path segments that are not covered due to obstacle avoidance deviation in real time and establishes a queue of missed sweeping tasks. After the global path is traversed, the robot is automatically scheduled to perform secondary fill-in work on the missed areas in the queue.

[0008] Furthermore, the improved DWA local planner in velocity space Internal sampling, with the sampling period set as follows: The robot's linear velocity at the current moment is angular velocity is After sampling and obtaining multiple candidate predicted trajectories, an improved trajectory evaluation function is used to score and rank each trajectory to select the optimal obstacle avoidance trajectory. The improved trajectory evaluation function is as follows: ; in, These are the smoothing normalization coefficients; The azimuth evaluation sub-function assesses the angular difference between the orientation of the predicted trajectory's end and the local target point. The safety distance evaluation subfunction assesses the distance between the predicted trajectory and the nearest obstacle; For velocity evaluation sub-functions; The newly added coverage and following evaluation subfunction is used to constrain the robot to follow the original path as closely as possible during obstacle avoidance. These are the weights of each evaluation sub-function, determined based on the task scenario, and must satisfy... .

[0009] Furthermore, the calculation formula for the coverage follow-up evaluation sub-function is as follows: ; Where k is the penalty gain constant, For sampling speed The shortest Euclidean distance from the end position of the predicted trajectory to the set of unexecuted global "bow" shaped paths.

[0010] Furthermore, the coverage follow-up evaluation sub-function A piecewise nonlinear penalty mechanism is adopted: when the lateral error Less than or equal to the safety threshold When the lateral error is high, a linear penalty function is used; when the lateral error is low, a linear penalty function is used. Greater than the safety threshold At this time, a sharp penalty function based on the inverse of the square of the distance is used to force the robot to prioritize reverting to the global reference path. The specific calculation formula is as follows: ; In the formula, and For smoothing adjustment coefficient, For sampling speed The shortest Euclidean distance from the end of the predicted trajectory to the set of unexecuted global "bow"-shaped paths.

[0011] Furthermore, when the robot deviates from its original global path during local obstacle avoidance, the system establishes an "environmental memory" mechanism to calculate the operating radius in real time. For any uncovered global path segments, establish a queue of missed scan tasks. Specifically, for any point on the original global path If it is consistent with the actual driving trajectory The shortest distance is always greater than the operating radius. Then, the coordinates of that point will be added to the missed scan task queue as a compensation waypoint. middle.

[0012] Furthermore, after the global path traversal is complete, the system automatically checks the missed scan task queue. ,right Secondary patching operation is performed on the missed areas in the scan queue: First, the missed scan task queue is checked. The discrete waypoints in the data are clustered using DBSCAN to form gap-filling blocks. These gap-filling blocks are then treated as nodes, and the LKH algorithm is used to find the shortest connection path to traverse all gap-filling blocks. Finally, when entering the gap-filling block, the local "bow" or "return" shaped coverage operation is restored, thereby achieving full coverage in complex dynamic environments.

[0013] A full-coverage path planning system for a sweeping and mowing robot includes an environmental perception module, a global path planning module, an improved DWA local planning module, a job execution and status management module, and a missed sweeping record and missed replenishment scheduling module. Environmental perception module: used to scan the working environment in real time, and obtain boundary information of the target area, static environment map, and the position and movement status of dynamic obstacles; Global path planning module: Receives boundary information and static occupied grid map from the environment perception module, generates a "bow"-shaped global path using the ox-plowing cell decomposition method, and sends it to the job execution and status management module as the main tracking path for the robot during normal operation; Improved DWA Local Planning Module: This module is triggered when the environment perception module detects an unknown obstacle. It samples in the velocity space and uses an improved trajectory evaluation function to evaluate the trajectory. The improved trajectory evaluation function adds a coverage following factor to constrain the lateral error between the obstacle avoidance trajectory and the unexecuted global path, and outputs the optimal obstacle avoidance trajectory. The task execution and status management module receives the global path from the global path planning module, the obstacle avoidance trajectory from the improved DWA local planning module, and the gap-filling connection path from the gap-scanning recording and gap-filling scheduling module. It controls the robot's movement according to the current task mode. At the same time, it records the actual driving trajectory and the covered area in real time, updates the execution index of the global path in real time, marks physically covered path points as "completed", and feeds back the gap-scanning recording and gap-filling scheduling module any uncovered path segments caused by obstacle avoidance. The omission scanning record and omission remediation scheduling module receives the actual trajectory and coverage data reported by the job execution and status management module. Based on the real-time occupied grid map, it makes a judgment: for points on the original global path that are not covered by the effective operation radius due to obstacle avoidance deviation, its coordinates are stored in the omission scanning task queue. After the global main task is completed, this module automatically checks the omission scanning task queue. If it is not empty, it executes the omission remediation path optimization strategy and generates an efficient omission remediation connection path, which is then sent to the job execution and status management module.

[0014] Furthermore, the environmental perception module uses LiDAR to acquire the distance contours of obstacles and combines it with wheeled odometers to calculate and locate them for real-time scanning of the work environment; during obstacle avoidance, the SLAM algorithm is used to construct a real-time occupied grid map, and the waypoint extraction of the missed scan task queue is based on the grid coordinates of the grid in the real-time occupied grid map that are in the state of "unknown" or "not covered by the work radius".

[0015] Furthermore, cover the following factors , used to constrain the lateral error between the obstacle avoidance trajectory and the unexecuted global path, where k is the penalty gain constant. For sampling speed The shortest Euclidean distance from the end position of the predicted trajectory to the set of unexecuted global "bow" shaped paths.

[0016] Furthermore, the specific strategy for optimizing the gap-filling path is as follows: First, for The discrete waypoints in the data are clustered using DBSCAN, and the missed scan points with a distance less than a set threshold are merged into "missed scan blocks". Then, the missed scan blocks are treated as nodes, and the LKH algorithm is used to solve the shortest connection path to traverse all missed scan blocks. Finally, the optimized missed scan connection path is sent to the job execution and status management module to generate "bow" shaped sub-paths inside the missed scan blocks. After the robot enters the missed scan blocks, it resumes the local "bow" shaped operation.

[0017] Beneficial effects: 1. This invention introduces a coverage following factor into the DWA evaluation function, giving the local obstacle avoidance process a "centripetal force" on the global path, which solves the problem of large-area missed scans caused by arbitrary deviation from the path during obstacle avoidance in traditional methods, and significantly improves the operational coverage during the obstacle avoidance process.

[0018] 2. This invention uses an improved evaluation function to force the selection of the trajectory with the smallest lateral error, achieving a smooth return to the original path after the obstacle avoidance action. This effectively avoids path oscillation and repetitive work caused by traditional PID control, improves work efficiency, and reduces hardware wear and tear.

[0019] 3. This invention establishes a queue of missed scan tasks to perform "environmental memory," transforming uncovered areas caused by obstacle avoidance into quantifiable compensation tasks. After the main task is completed, secondary gap filling is performed, ensuring that the final full coverage rate reaches the theoretical value from the system level.

[0020] In summary, this invention achieves full-coverage path planning by integrating environmental memory and an improved DWA algorithm, effectively reducing the missed area and regression oscillation during obstacle avoidance, significantly improving the full coverage rate and operational smoothness in complex dynamic environments, reducing hardware wear and tear, and is suitable for full-coverage path planning of various mobile cleaning and lawn mowing robots. Attached Figure Description

[0021] Figure 1 This is a system module architecture diagram provided for an embodiment of the present invention.

[0022] Figure 2 The main flowchart of the full-coverage path planning algorithm provided in the embodiments of the present invention is shown.

[0023] Figure 3 This is a schematic diagram illustrating the principle of local obstacle avoidance and coverage following in an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram comparing the obstacle avoidance trajectory and lateral offset provided in an embodiment of the present invention.

[0025] Figure 5 This is a comparison chart of missed scan areas provided in an embodiment of the present invention.

[0026] Figure 6 This is a task coverage comparison chart provided for an embodiment of the present invention. Detailed Implementation

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0028] like Figure 1 The diagram shown is a module architecture diagram of the full-coverage path planning system for the lawn mowing robot in this embodiment. The system mainly includes an environmental perception module, a global path planning module, an improved DWA local planning module, a job execution and status management module, and a missed sweeping record and missed replenishment scheduling module.

[0029] Environmental perception module: Equipped with single-line or multi-line LiDAR, depth camera and other airborne sensors, and combined with the chassis wheel odometer, it is used to scan the working environment in real time; The system uses SLAM algorithms such as Gmapping or Cartographer to build a real-time occupied grid map, and obtains the boundary information of the target area, the known static environment map and the position and movement status of dynamic obstacles.

[0030] Global Path Planning Module: This module receives boundary information and a static occupancy grid map from the environmental perception module. It uses a slack-scale cell decomposition method to divide the target work area into several unobstructed sub-regions. Within each sub-region, a "bow"-shaped global path is generated based on the robot's effective work radius, ensuring that the path spacing matches the work width and achieving theoretically complete coverage. The generated global path point sequence is sent to the job execution and status management module as the main tracking reference for the robot during normal operation. When the job execution module reports that the global main task has been completed, this global path planning module enters standby mode until a secondary gap-filling task is triggered. If it is necessary to replan the coverage path within a local gap-filling block, this module, in conjunction with the gap-scanning recording and gap-filling scheduling module, generates a "bow"-shaped sub-path within the gap-filling block.

[0031] Improved DWA Local Planning Module: This module is triggered when the environment perception module detects an unknown obstacle. The improved DWA local planning module samples within the velocity space and uses an improved evaluation function to evaluate the trajectory. This improved evaluation function adds a coverage following factor. , used to constrain the lateral error between the obstacle avoidance trajectory and the unexecuted global path, where k is the penalty gain constant. For sampling speed The module calculates the shortest Euclidean distance from the end position of the predicted trajectory to the set of unexecuted global "bow" shaped paths; it outputs the optimal obstacle avoidance trajectory.

[0032] Job Execution and Status Management Module: This module serves as the system's execution center. It receives the global path from the global path planning module, the obstacle avoidance trajectory from the improved DWA local planning module, and the missing path from the missed scan recording and missing path scheduling module. It then controls the robot's movement according to the current job mode (global tracking mode, local obstacle avoidance mode, or missing path scheduling mode). Simultaneously, this module records the actual driving trajectory and covered areas in real time, updates the execution index of the global path, marks physically covered path points as "completed," and reports any uncovered path segments caused by obstacle avoidance to the missed scan recording and missing path scheduling module.

[0033] The omission tracking and remediation scheduling module receives the actual trajectory and coverage data reported by the job execution and status management module. This module determines the omission based on the real-time occupied grid map: for points on the original global path that are not covered by the effective work radius due to obstacle avoidance deviation, their coordinates are stored in the omission task queue. After the global main task is completed, the omission tracking and remediation scheduling module automatically checks the omission task queue. If it is not empty, it executes a remediation path optimization strategy: first, it performs DBSCAN clustering on the discrete waypoints in the omission task queue, merging omission points with a distance less than a set threshold (1m in this embodiment) into "remediation blocks"; then, it treats these remediation blocks as Traveling Salesman Problem (TSP) nodes, using the LKH algorithm to solve for the shortest connecting path to traverse all remediation blocks; finally, it sends the optimized remediation connecting path to the job execution and status management module. After entering the remediation block, the robot resumes its local "bow"-shaped operation, avoiding unnecessary round trips during the secondary remediation process.

[0034] The five modules are connected and interact as follows: The environmental perception module sends boundary and map data to the global path planning module and sends obstacle trigger signals to the improved DWA local planning module; the global path planning module sends the generated bow-shaped global path to the job execution and status management module; the improved DWA local planning module sends the optimal obstacle avoidance trajectory to the job execution and status management module; the job execution and status management module sends the actual trajectory and information on uncovered points to the missed scan recording and gap filling scheduling module; when the global main task is completed, the missed scan recording and gap filling scheduling module sends the gap filling path to the job execution and status management module.

[0035] Based on the aforementioned full-coverage path planning system for sweeping and mowing robots, this invention also provides a method for full-coverage path planning for sweeping and mowing robots. This method first uses a ox-plowing-style unit decomposition method to generate a "bow"-shaped global reference path. During the robot's operation along the global path, when the sensor detects an unknown obstacle, an improved DWA local planner is triggered. A coverage following factor is added to its evaluation function to minimize the lateral error between the obstacle avoidance trajectory and the global path, achieving a smooth transition between obstacle avoidance and tracking. Simultaneously, the system records path segments not covered due to obstacle avoidance deviation in real time, establishing a queue of missed sweeping tasks. After the global main path has been traversed, the robot is automatically scheduled to perform secondary fill-in operations on the missed areas in the queue.

[0036] like Figure 2 As shown, the algorithm execution flow of a full-coverage path planning method for a cleaning and mowing robot is as follows, including steps S1-S8.

[0037] S1. Start the task and obtain the boundary and static map: The robot starts the task and scans the current environment through onboard sensors (such as LiDAR, depth camera, etc.) to obtain the boundary information of the target work area and the known static environment map.

[0038] S2. Generate a "bow"-shaped global operation path: At the global planning layer, the system uses a slack farming-style unit decomposition method to divide the target operation area into multiple barrier-free sub-regions. Subsequently, a "bow"-shaped global path with high coverage is generated within each sub-region.

[0039] S3. Start the operation and travel along the global path: The robot begins to perform cleaning or mowing operations along the planned "bow"-shaped global path. During this process, the system continuously monitors whether the global path has been traversed.

[0040] S4, Detecting Unknown Obstacles: While moving along the original path, the robot uses sensors to perceive the environment in real time and determine whether there are unknown static or dynamic obstacles ahead. If no obstacle is detected (No): Enter S4.1, the system updates the tracking index of the global path according to the robot's current pose, and marks the traversed path points as "operated"; then loop back to S3, driving the robot to continue to travel smoothly along the global path; If an obstacle is detected: trigger the local obstacle avoidance mechanism, and enter S5.

[0041] S5. Trigger the improved DWA and calculate the optimal obstacle avoidance trajectory: When an obstacle is detected, the system triggers the improved dynamic window method (improved DWA local planner) proposed in this invention to perform local obstacle avoidance. The specific method is as follows: Kinematic Model Construction and Velocity Sampling: The improved dynamic window method for velocity space sampling is based on the robot's two-wheel differential kinematic model. Let the sampling period be... The robot's linear velocity at the current moment is angular velocity is Then the iterative calculation formula for predicting trajectory coordinates is: ; By iteratively calculating using the above formula, multiple candidate predicted trajectories are obtained. The velocity sampling range is subject to physical and electromechanical constraints on the robot's maximum acceleration / deceleration and minimum turning radius.

[0042] Improved trajectory evaluation: in velocity space After obtaining multiple predicted trajectories through mid-sample, an improved trajectory evaluation function is used for scoring and ranking. The improved trajectory evaluation function is defined as follows: ; in, These are the smoothing normalization coefficients; The azimuth evaluation sub-function assesses the angular difference between the orientation of the predicted trajectory's end and the local target point. The safety distance evaluation subfunction assesses the distance between the predicted trajectory and the nearest obstacle; The speed evaluation subfunction encourages the robot to maintain a high operating speed while ensuring safety. The newly added coverage and following evaluation subfunction is used to constrain the robot to follow the original path as closely as possible during obstacle avoidance. These are the weights of each evaluation sub-function, determined based on the task scenario, and must satisfy... .

[0043] In this embodiment, the parameters of the evaluation function are calibrated as follows: the normalization constraint of the evaluation function weight coefficients satisfies... In typical open-air work scenarios, the recommended settings are as follows: =0.2, =0.3, =0.1, =0.4; In crowded scenarios where the sensor detects numerous obstacles, the system adaptively and dynamically adjusts the safety distance weight. To ensure absolute safety, the effective operating radius should be above 0.5. The dimensions are determined based on the actual chassis size of the robot for mowing / sweeping, and the preferred size is: the physical working width of the robot body + 10cm overlap redundancy.

[0044] The typical method for calculating the coverage follow-through evaluation sub-function is as follows: ; In the formula, k is the penalty gain constant, and Δd(v,ω) is the shortest Euclidean distance from the end position of the predicted trajectory to the unexecuted global "bow" shaped path set at the sampling velocity (v,ω).

[0045] To more effectively suppress long-distance derailment during obstacle avoidance, in this preferred embodiment, the coverage following evaluation sub-function employs a piecewise nonlinear penalty mechanism: when the lateral error Δd(v,ω) is less than or equal to the safety threshold... When the value of Coverage remains at a high level, when Δd(v,ω) exceeds... At this point, the Coverage value drops sharply to force the robot to prioritize the return path. The specific calculation formula is as follows: In the formula, and The smoothing adjustment coefficient is Δd(v,ω), which is the shortest Euclidean distance from the end of the predicted trajectory to the set of unexecuted global "bow"-shaped paths.

[0046] like Figure 3 As shown, by using the improved trajectory evaluation function, the system forces the selection of the trajectory with the smallest lateral error, thereby achieving a smooth transition between obstacle avoidance and line following.

[0047] S6. Perform obstacle avoidance and push missed points into the queue: The robot performs local obstacle avoidance according to the optimal trajectory selected in S5. During the deviation from the original path, the system will establish an "environment memory" mechanism to push the missed points into the queue.

[0048] The "environmental memory" mechanism specifically involves the system calculating in real time the global path segments not covered by the sweeping or mowing operation radius; assuming the robot's effective operating radius is... The actual driving trajectory is For any point on the original global path If the following conditions are met Then its coordinates are added to the missed scan task queue as compensation waypoints. middle.

[0049] S7. Determine if the global path has been re-attached: The system detects whether the obstacle avoidance action has been completed and determines whether the robot's current position has smoothly returned to the original "bow"-shaped global path.

[0050] If the obstacle is not aligned (no), return to step S5 and continue to calculate the optimal obstacle avoidance and alignment trajectory for the next moment; if the obstacle is aligned, proceed to S7.1.

[0051] S7.1 End the obstacle avoidance process, switch back to global tracking mode, and then proceed to step S4.1 to resynchronize the path execution progress; the system calculates the projection point of the current pose on the global path, skips the expired path points that have been bypassed during obstacle avoidance, and resets the tracking target to the nearest pending work point in front, so as to continue to execute the main task; before the global path traversal is completed, repeat S4-S7; if the global path traversal is completed, proceed to step S8 to make a check for omissions.

[0052] S8. Check for omissions after global path traversal: After the global main task (i.e., all planned paths when there are no obstacles) is completed, the system determines whether to perform a second omission check operation and whether the omission scan queue is empty.

[0053] If the missed scan queue is not empty (No): Trigger a missed scan operation. The system schedules the robot to use global pathfinding algorithms such as A* to plan the path to the missed scan area, and sequentially checks the missed scan queue. The missing areas in the process are subjected to secondary missing area repair operations; among them, the LKH algorithm is used to determine the optimal access order of multiple missing area blocks, while when actually moving between blocks, the A* algorithm is still used to plan a collision-free global path.

[0054] If the missed scan queue is empty (yes) or the missing scan operation has been completed: This signifies that the 100% full coverage task in a complex dynamic environment has been truly completed. The robot finishes its full-map operation and automatically returns to the charging dock.

[0055] Comparison Test of the Effects of Embodiments of the Invention To further verify the technical effect of the present invention, in a 100-square-meter ROS2 simulated lawn environment with 5 randomly distributed dynamic / static obstacles, robots with the same chassis kinematic parameters were tested 20 times each using the traditional DWA scheme and the scheme of fusion environment memory and improved DWA disclosed in the present invention. The statistical average indicators are shown in Table 1 below.

[0056] Table 1 Statistical evaluation indicators of the present invention and traditional DWA schemes Experimental data conclusively demonstrate that this invention effectively solves the path oscillation and large-area missed sweep blind zone problems caused by traditional local obstacle avoidance algorithms by introducing a coverage following factor and a segmented penalty mechanism. At the same time, in conjunction with the DBSCAN and LKH optimized omission-filling scheduling module, it achieves almost perfect job coverage in complex dynamic environments, and has extremely high engineering practical value and significant progress.

[0057] Figure 4This is a comparison test of obstacle avoidance trajectory and lateral offset in a ROS2 simulation environment. The dotted line represents the ideal global "bow"-shaped reference path, the dashed line represents the obstacle avoidance trajectory of the traditional DWA algorithm, and the thick solid line represents the obstacle avoidance trajectory of the improved algorithm of this invention. It can be seen that the traditional algorithm, after avoiding the obstacle, generates a lateral offset as high as 0.46m due to the lack of path constraints, and oscillates during the regression process, resulting in a large missed area behind the obstacle. However, using the solution of this invention, under the strong constraint of the coverage following evaluation factor, the trajectory is quickly and smoothly pulled back to the global reference path, and the maximum lateral offset is significantly compressed to 0.12m, effectively avoiding long-distance derailment.

[0058] Figure 5 ,and Figure 6 The images show a direct comparison of the two schemes in terms of core operational indicators. Figure 5 The comparison chart of missed scan areas shows that the average missed scan area per scan of the present invention is only 0.18m². 2 Compared to the traditional solution of 0.87m 2 A significant reduction of 79.3% was achieved; Figure 6 The coverage comparison chart further shows that, thanks to the establishment of the missed sweep queue and the secondary missing sweep scheduling mechanism, the final full map coverage of the present invention jumps from 88.5% of the traditional scheme to 99.8%, greatly approaching the theoretical full coverage limit of 100%, and significantly improving the working efficiency of the sweeping and mowing robot in complex dynamic environments.

[0059] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for full-coverage path planning for a lawn mowing robot, characterized in that, First, a "bow"-shaped global path is generated using the ox-plowing-style unit decomposition method. When the robot operates along the global path, the improved DWA local planner is triggered when the sensor detects an unknown obstacle. The trajectory evaluation function adds a coverage following factor to minimize the lateral error between the obstacle avoidance trajectory and the global path, and outputs the optimal obstacle avoidance trajectory to achieve a smooth transition between obstacle avoidance and line following. At the same time, the system records the path segments that are not covered due to obstacle avoidance deviation in real time and establishes a queue of missed scan tasks. After the global path is traversed, the robot is automatically scheduled to perform secondary fill-in work on the missed areas in the queue.

2. The method for full-coverage path planning of a lawnmower robot according to claim 1, characterized in that, The improved DWA local planner in velocity space Internal sampling, with the sampling period set as follows: The robot's linear velocity at the current moment is angular velocity is After sampling and obtaining multiple candidate predicted trajectories, an improved trajectory evaluation function is used to score and rank each trajectory to select the optimal obstacle avoidance trajectory. The improved trajectory evaluation function is as follows: ; in, These are the smoothing normalization coefficients; The azimuth evaluation sub-function assesses the angular difference between the orientation of the predicted trajectory's end and the local target point. The safety distance evaluation subfunction assesses the distance between the predicted trajectory and the nearest obstacle; For velocity evaluation sub-functions; The newly added coverage and following evaluation subfunction is used to constrain the robot to follow the original path as closely as possible during obstacle avoidance. These are the weights of each evaluation sub-function, determined based on the task scenario, and must satisfy... .

3. The method for full-coverage path planning of a lawnmower robot according to claim 2, characterized in that, The calculation formula for the coverage follow-up evaluation sub-function is as follows: ; Where k is the penalty gain constant, For sampling speed The shortest Euclidean distance from the end position of the predicted trajectory to the set of unexecuted global "bow" shaped paths.

4. The method for full-coverage path planning of a lawnmower robot according to claim 2, characterized in that, The coverage follow-up evaluation subfunction A piecewise nonlinear penalty mechanism is adopted: when the lateral error Less than or equal to the safety threshold When the horizontal error is 1, a linear penalty function is used; when the horizontal error is 1, a linear penalty function is used. Greater than the safety threshold At this time, a sharp penalty function based on the inverse of the square of the distance is used to force the robot to prioritize reverting to the global reference path. The specific calculation formula is as follows: ； In the formula, and For smoothing adjustment coefficient, For sampling speed The shortest Euclidean distance from the end of the predicted trajectory to the set of unexecuted global "bow"-shaped paths.

5. The method for full-coverage path planning of a lawnmower robot according to claim 1, characterized in that, When the robot deviates from its original global path during local obstacle avoidance, the system establishes an "environmental memory" mechanism to calculate the operating radius in real time. For any uncovered global path segments, establish a queue of missed scan tasks. Specifically, for any point on the original global path If it is consistent with the actual driving trajectory The shortest distance is always greater than the operating radius. Then, the coordinates of that point will be added to the missed scan task queue as a compensation waypoint. middle.

6. The method for full-coverage path planning of a lawnmower robot according to claim 5, characterized in that, After the global path traversal is complete, the system automatically checks the missed scan task queue. ,right Secondary patching operation is performed on the missed areas in the scan queue: First, the missed scan task queue is checked. The discrete waypoints in the data are clustered using DBSCAN to form gap-filling blocks. These gap-filling blocks are then treated as nodes, and the LKH algorithm is used to find the shortest connection path to traverse all gap-filling blocks. Finally, when entering the gap-filling block, the local "bow" or "return" shaped coverage operation is restored, thereby achieving full coverage in complex dynamic environments.

7. A full-coverage path planning system for a lawn mowing robot, characterized in that, It includes an environmental perception module, a global path planning module, an improved DWA local planning module, a job execution and status management module, and a missed scan recording and missing scan scheduling module. Environmental perception module: used to scan the working environment in real time, and obtain boundary information of the target area, static environment map, and the position and movement status of dynamic obstacles; Global path planning module: Receives boundary information and static occupied grid map from the environment perception module, generates a "bow"-shaped global path using the ox-plowing cell decomposition method, and sends it to the job execution and status management module as the main tracking path for the robot during normal operation; Improved DWA Local Planning Module: This module is triggered when the environment perception module detects an unknown obstacle. It samples in the velocity space and uses an improved trajectory evaluation function to evaluate the trajectory. The improved trajectory evaluation function adds a coverage following factor to constrain the lateral error between the obstacle avoidance trajectory and the unexecuted global path, and outputs the optimal obstacle avoidance trajectory. The task execution and status management module receives the global path from the global path planning module, the obstacle avoidance trajectory from the improved DWA local planning module, and the gap filling connection path from the gap scanning record and gap filling scheduling module. It controls the robot's movement according to the current task mode. At the same time, it records the actual driving trajectory and the covered area in real time, updates the execution index of the global path in real time, marks physically covered path points as "completed", and feeds back the gap scanning record and gap filling scheduling module any uncovered path segments caused by obstacle avoidance. The omission scanning record and omission remediation scheduling module receives the actual trajectory and coverage data reported by the job execution and status management module. Based on the real-time occupied grid map, it makes a judgment: for points on the original global path that are not covered by the effective operation radius due to obstacle avoidance deviation, its coordinates are stored in the omission scanning task queue. After the global main task is completed, this module automatically checks the omission scanning task queue. If it is not empty, it executes the omission remediation path optimization strategy and generates an efficient omission remediation connection path, which is then sent to the job execution and status management module.

8. A full-coverage path planning system for a lawn mowing robot according to claim 7, characterized in that, The environmental perception module uses LiDAR to acquire the distance contours of obstacles and combines it with wheeled odometers to calculate and locate them for real-time scanning of the working environment. During obstacle avoidance, the SLAM algorithm is used to construct a real-time occupied grid map. The waypoint extraction of the missed scan task queue is based on the grid coordinates of grids in the real-time occupied grid map that are in the state of "unknown" or "not covered by the working radius".

9. A full-coverage path planning system for a lawn mowing robot according to claim 7, characterized in that, Coverage Follow-up Factor This is used to constrain the lateral error between the obstacle avoidance trajectory and the unexecuted global path, where k is the penalty gain constant. For sampling speed The shortest Euclidean distance from the end position of the predicted trajectory to the set of unexecuted global "bow" shaped paths.

10. A full-coverage path planning system for a lawn mowing robot according to claim 7, characterized in that, The specific strategy for optimizing the leak detection path is as follows: First, optimize the leak detection task queue. DBSCAN clustering is performed on discrete waypoints in the data, and missed scan points with a distance less than a set threshold are merged into "missed scan blocks"; The patching block is then treated as a node, and the LKH algorithm is used to solve the shortest connection path for traversing all patching blocks. Finally, the optimized patching connection path is sent to the job execution and status management module to generate a "bow"-shaped sub-path inside the patching block. After the robot enters the patching block, it resumes the local "bow"-shaped operation.

Images