Self-learning obstacle avoidance control method for mowing robot

Abstract

A self-learning obstacle avoidance control method for a mowing robot. The method comprises the following steps: S1, planning a global path, involving: acquiring position data of a current lawn by means of a preset processor and a positioning module, downloading an online map from a wide area network on the basis of the position data acquired by the processor, and planning an optimal mowing path by means of an algorithm; S2, performing adaptive obstacle avoidance, involving: determining a shape feature of an obstacle on the basis of a motion parameter of a mowing robot itself during the execution of a task, and performing a mowing task along the edge of the obstacle; S3, updating the mowing path in real time, involving: marking the position of the obstacle encountered in the mowing task; and S4, planning a supplementary mowing path, involving: setting as a target region the region that has not been subjected to mowing during the mowing task, and recording the position of the target region. By means of steps S1-S4 provided in the method, the mowing repetition rate is reduced, such that the mowing area coverage is uniform, thereby improving the mowing effect.

Description

Self-learning obstacle avoidance control method for lawnmower robots Technical Field

[0001] This invention relates to the field of lawnmower technology, and in particular to a self-learning obstacle avoidance control method for lawnmowers. Background Technology

[0002] Currently, most commercially available intelligent mobile lawnmowers employ a method of establishing an electronic fence by embedding electrified wires along the boundary, then identifying the boundary through electromagnetic induction. Within the wire boundary, most lawnmowers move along random paths while using distance sensors to avoid obstacles at appropriate distances. A few advanced models use a full-coverage, plow-like path planning approach. After completing the mowing task, the lawnmower typically returns to its charging station along the boundary.

[0003] However, existing lawnmower robots still have the following problems:

[0004] (1) In the existing movement mode of lawn mowing robots, the movement trajectory of random paths is often not purposeful. When encountering obstacles and changing direction, it is very likely to cause problems such as extended working time, change of mowing direction, and increased mowing repetition rate, which can easily lead to uneven coverage of mowing area and thus fail to achieve the effect of complete mowing.

[0005] (2) For obstacles in the lawn, wires need to be added to delineate restricted areas, which increases the complexity of the map, the amount of wire laying work and the cost. In addition, the path of the energized wires that must be added to the wires will divide the blank map, so that the grass in this area will be ignored by the mowing robot.

[0006] Therefore, we propose a self-learning obstacle avoidance control method for lawnmower robots. Summary of the Invention

[0007] The main objective of this invention is to provide a self-learning obstacle avoidance control method for lawnmower robots. By setting up real-time path planning, adaptive obstacle avoidance, and real-time updating of the mowing path, this invention addresses the problems in existing lawnmower robot movement methods. These methods often involve random paths with no clear purpose, leading to increased working time, changes in mowing direction, and increased mowing repetition rates when encountering obstacles. This can result in uneven mowing coverage and incomplete mowing. The invention also addresses the issue of adding wires to demarcate restricted areas around obstacles in the lawn, increasing map complexity, wiring work, and costs. Furthermore, the added wire paths can segment blank areas on the map, causing the lawnmower robot to ignore these areas.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] The self-learning obstacle avoidance control method for lawnmower robots includes the following steps:

[0010] S1. Global path planning: The preset processor and positioning module obtain the current location data of the grass, and the preset mobile terminal establishes a communication connection with the processor through the local area network communication module. Based on the location data obtained by the processor, the online map is downloaded from the wide area network. The operator transmits the boundary information of the mowing area to the processor through the mobile terminal, and plans the optimal mowing path through the algorithm.

[0011] S2. Adaptive obstacle avoidance: After receiving the instructions from the mobile terminal, the lawnmower robot performs the lawnmower task according to the optimal lawnmower path. During the task execution, it judges the shape characteristics of the obstacle based on the lawnmower robot's own motion parameters and performs the lawnmower task along the edge of the obstacle.

[0012] S3. Real-time update of mowing path: Mark the location of obstacles encountered in the mowing task and update the mowing path in real time.

[0013] S4. Path planning for additional mowing: Set the un-mowed areas along the mowing task as target areas and record the location of the target areas. When the mowing robot performs the mowing task according to the mowing path, it judges the distance between the location of the target area and the updated path and further optimizes the mowing path.

[0014] Preferably, the acquisition of the mowing area boundary information in step S1 includes the following steps:

[0015] S11. Preset a drone and use the drone's onboard downward-looking binocular passive vision system to traverse the grassland boundary and generate a plot boundary coordinate map;

[0016] S12. Based on the land parcel boundary coordinate map, identify obstacles within the land parcel boundary and correct the land parcel boundary map;

[0017] S13. Visual recognition and segmentation technology is used to divide the corrected plot boundaries, determine the drivable area and obstacle boundary information, and plan the optimal drivable area path information.

[0018] Preferably, the algorithm in step S1 is a genetic algorithm.

[0019] Preferably, determining the shape features of the obstacle in step S2 includes the following steps:

[0020] S21. A preset CCD camera acquires omnidirectional image features of the obstacle, and a preset processing module divides the obstacle image into n equal regions, setting the color pixels of each small region to be... The number of edge pixels is obtained after edge detection. The processing module combines color pixels and edge pixels into a feature set.

[0021] S22. Further judgment is made based on the ratio of the number of pixels detected for edge and color. If the ratio is within a set threshold, the area is determined to be an obstacle, as shown in the following formula:

[0022] ;

[0023] In the formula, α and β are the upper and lower limits of the ratio between the two; This is the threshold for the number of colors in a region.

[0024] Preferably, the task of mowing along the edge of the obstacle in step S2 includes the following steps:

[0025] S201. The robot detects obstacles in the grass by using the laser radar and CCD camera preset in the lawnmower, and retrieves the corresponding mowing strategy preset in the lawnmower database through the preset control unit. Then, the lawnmower strategy is converted into instructions and transmitted to the receiving unit of the lawnmower through the preset transmission unit.

[0026] S202. Control the robot to avoid obstacles and perform grass mowing operations by using a grass-mowing strategy.

[0027] Preferably, the mowing strategy in step S201 involves using the laser radar and CCD camera of the mowing robot to capture and record obstacles in real time, reconstructing the obstacle model using a preset measurement method, and then calculating the optimal mowing path of the mowing robot using a genetic algorithm.

[0028] Preferably, marking the locations of obstacles encountered during the mowing task in step S3 includes the following steps:

[0029] S2021. Generate a path map based on the trajectory of the lawnmower robot around the obstacles, and generate corresponding obstacle labels, and store the labels in a preset memory.

[0030] S2022, The control unit retrieves obstacle tags from the memory and establishes a connection between the tag attributes and the mowing strategy, thus establishing an association.

[0031] Preferably, the target area setting in step S4 includes the following steps:

[0032] S41. When the lawnmower robot performs lawnmowing tasks along the mowing path, it scans the grass around the path in real time using a CCD camera, determines the unmowed grass area based on the grass height, and records the coordinates and shape of this area to generate the main area node.

[0033] S42. Radialize outwards from the main region node, scan the uncut grassland adjacent to the main region node, record its coordinates and shape, and generate sub-region nodes.

[0034] S43. Establish the relationship between the main region nodes and the sub-region nodes, and generate the target region mesh map based on the distribution of the main region nodes and the sub-region nodes. Calculate the optimal cutting path connecting the main region nodes and multiple sub-region nodes using a genetic algorithm.

[0035] Preferably, the step S4 of determining the distance between the target area location and the updated path includes the following steps:

[0036] S401. Obtain the first target area to be processed near the mowing path of the lawn mowing robot, control the lawn mowing robot to move along the mowing path according to the movement command set by the user, and obtain the positioning information of the lawn mowing robot in real time during the movement. Based on the positioning information of the lawn mowing robot, determine the first target area to be processed.

[0037] S402: Expand the first target area to be processed outward by a preset distance to become the second target area to be processed, and control the lawn mowing robot to move along the second target area to be processed.

[0038] S403. Obtain environmental information of the lawnmower in the second target area based on the sensors of the lawnmower, determine the boundary of the non-working area in the environmental information, control the lawnmower to move along the boundary of the non-working area to obtain the third target area to avoid the non-working area, and construct the boundary of the working area of ​​the lawnmower based on the second target area, the third target area and the subsequent target areas.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] I. In this invention, by setting up real-time path planning, adaptive obstacle avoidance, and real-time updating of the mowing path, the optimal mowing path for the grassland is first planned. Then, based on the shape characteristics of the obstacles, the mowing path is replanned. A mowing strategy is formulated to control the robot to avoid obstacles and perform mowing operations. Furthermore, through the connection between tags and mowing strategies, the corresponding mowing strategy can be easily retrieved through the tags, saving retrieval time, reducing the mowing repetition rate, making the mowing area evenly covered, and improving the mowing effect.

[0041] Second, in this invention, by setting up a supplementary mowing path planning, the area of ​​grass that has not been mowed during the mowing task is set as the target area, and the position of the target area is recorded. When the mowing robot performs the mowing task according to the mowing path, it judges the distance between the target area and the updated path, and further optimizes the mowing path, thereby supplementing the missed parts of the target area near the obstacle. There is no need to bury wires to divide the restricted area, which further improves the mowing efficiency and mowing effect. Attached Figure Description

[0042] Figure 1 is a flowchart of the self-learning obstacle avoidance control method of the lawnmower robot of the present invention;

[0043] Figure 2 is a flowchart of the process for obtaining the boundary information of the mowing area in this invention;

[0044] Figure 3 is a flowchart of the obstacle location marking process encountered in the lawn mowing task of this invention;

[0045] Figure 4 is a flowchart of the target area setting process of the present invention;

[0046] Figure 5 is a flowchart illustrating the process of determining the distance between the target area and the updated path according to the present invention. Detailed Implementation

[0047] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0048] Example: As shown in Figures 1-5, the self-learning obstacle avoidance control method for a lawnmower robot includes the following steps:

[0049] S1. Global path planning: The preset processor and positioning module obtain the current location data of the grass, and the preset mobile terminal establishes a communication connection with the processor through the local area network communication module. Based on the location data obtained by the processor, the online map is downloaded from the wide area network. The operator transmits the boundary information of the mowing area to the processor through the mobile terminal, and the optimal mowing path is planned through the algorithm.

[0050] S2. Adaptive obstacle avoidance: After receiving instructions from the mobile terminal, the lawnmower robot performs the lawnmower task according to the optimal lawnmower path. During the task, it judges the shape characteristics of obstacles based on its own motion parameters and performs the lawnmower task along the edge of the obstacle.

[0051] S3. Real-time update of mowing path: Mark the location of obstacles encountered in the mowing task and update the mowing path in real time.

[0052] S4. Replenishment path planning: Set the uncut areas along the mowing task as target areas and record the location of the target areas. When the mowing robot performs the mowing task according to the mowing path, it judges the distance between the target area and the updated path and further optimizes the mowing path.

[0053] The acquisition of the mowing area boundary information in step S1 includes the following steps:

[0054] S11. Preset the drone and use the drone's onboard downward binocular passive vision system to traverse the grassland boundary and generate a plot boundary coordinate map.

[0055] S12. Based on the land parcel boundary coordinate map, identify obstacles within the land parcel boundary and correct the land parcel boundary map;

[0056] S13. Visual recognition and segmentation technology is used to divide the corrected plot boundaries, determine the drivable area and obstacle boundary information, and plan the optimal drivable area path information.

[0057] Obstacles include their number, category, and GPS coordinates;

[0058] The land parcel boundaries are traversed using an airborne binocular passive vision system on a drone to generate a land parcel boundary coordinate map, specifically including:

[0059] (1) Select the land boundary, obtain the rough GPS coordinates of the land boundary and the connection relationship between the coordinate points, store them as a doubly linked list data structure, and generate a rough land boundary coordinate map of the land boundary; the rough land boundary coordinate map includes the rough GPS coordinates, the connection relationship between the coordinate points and the doubly linked list data structure.

[0060] (2) Input the plot boundary coordinate map into the UAV, starting from the first node, select the direction of the next node as the basis for the plot boundary route, and establish a flight mission to traverse the plot boundary.

[0061] (3) During the flight mission, the UAV’s onboard downward binocular passive vision system continuously collects ground images and uses a deep learning network to determine the visual boundaries of the land plots; the visual boundaries of the land plots include field ridges, ditches, paved roadsides and water bodies;

[0062] (4) Combining the boundary points of the visual boundary of the land parcel with the relative position of the UAV, and based on the GPS coordinates of the UAV, assign GPS coordinates to the boundary points of the visual boundary of the land parcel.

[0063] (5) Update the doubly linked list data structure based on the boundary points, GPS coordinates and connection relationships between the new plot visual boundaries to generate an accurate plot boundary coordinate map.

[0064] The algorithm in step S1 is a genetic algorithm.

[0065] The step S2, determining the shape characteristics of the obstacle, includes the following steps:

[0066] S21. A preset CCD camera acquires omnidirectional image features of the obstacle, and a preset processing module divides the obstacle image into n equal regions, setting the color pixels of each small region to be... The number of edge pixels is obtained after edge detection. The processing module combines color pixels and edge pixels into a feature set.

[0067] S22. Further judgment is made based on the ratio of the number of pixels detected for edge and color. If the ratio is within a set threshold, the area is determined to be an obstacle, as shown in the following formula:

[0068] ;

[0069] In the formula, α and β are the upper and lower limits of the ratio between the two; This is the threshold for the number of colors in a region.

[0070] Obstacle images are segmented using the HS1 model and detected using the Canny edge detection operator to achieve fast and efficient obstacle edge recognition. To reduce computation time, images captured by the CCD camera are compressed using the Gussian algorithm.

[0071] The task of mowing the lawn along the edge of the obstacle in step S2 includes the following steps:

[0072] S201. The robot detects obstacles in the grass by using the laser radar and CCD camera preset in the lawnmower, and retrieves the corresponding mowing strategy preset in the lawnmower database through the preset control unit. Then, the lawnmower strategy is converted into instructions and transmitted to the receiving unit of the lawnmower through the preset transmission unit.

[0073] S202. Control the robot to avoid obstacles and perform grass mowing operations by using a grass-mowing strategy.

[0074] In step S201, the mowing strategy uses the laser radar and CCD camera of the mowing robot to take real-time pictures and record the obstacles, reconstructs the obstacle model through a preset measurement method, and then calculates the optimal mowing path of the mowing robot through a genetic algorithm.

[0075] The step S3, marking the locations of obstacles encountered in the lawn mowing task, includes the following steps:

[0076] S2021. Generate a path map based on the trajectory of the lawnmower robot around the obstacles, and generate corresponding obstacle labels, and store the labels in a preset memory.

[0077] S2022, The control unit retrieves obstacle tags from the memory and establishes a connection between the tag attributes and the mowing strategy, thus establishing an association.

[0078] The target area setting in step S4 includes the following steps:

[0079] S41. When the lawnmower robot performs lawnmowing tasks along the mowing path, it scans the grass around the path in real time using a CCD camera, determines the unmowed grass area based on the grass height, and records the coordinates and shape of this area to generate the main area node.

[0080] S42. Radialize outwards from the main region node, scan the uncut grassland adjacent to the main region node, record its coordinates and shape, and generate sub-region nodes.

[0081] S43. Establish the relationship between the main region nodes and the sub-region nodes, and generate the target region mesh map based on the distribution of the main region nodes and the sub-region nodes. Calculate the optimal cutting path connecting the main region nodes and multiple sub-region nodes using a genetic algorithm.

[0082] The recording module in the lawnmower records the starting coordinates of the first sub-region and the ending coordinates of the last sub-region within the same target area. The mowing planning device can create a mowing path for the sub-region based on the starting and ending coordinates of the sub-region, so that the mowing path can completely cover the unmowed area and minimize the mowing path within the sub-region. The lawnmower is then controlled to perform mowing operations on the target lawn within the sub-region along the mowing path, ensuring the mowing effect within the sub-region and improving the mowing efficiency.

[0083] The step S4, determining the distance between the target area and the updated path, includes the following steps:

[0084] S401. Obtain the first target area to be processed near the mowing path of the lawn mowing robot, control the lawn mowing robot to move along the mowing path according to the movement command set by the user, and obtain the positioning information of the lawn mowing robot in real time during the movement. Based on the positioning information of the lawn mowing robot, determine the first target area to be processed.

[0085] S402: Expand the first target area to be processed outward by a preset distance to become the second target area to be processed, and control the lawn mowing robot to move along the second target area to be processed.

[0086] S403. Obtain environmental information of the lawnmower in the second target area based on the sensors of the lawnmower, determine the boundary of the non-working area in the environmental information, control the lawnmower to move along the boundary of the non-working area to obtain the third target area to avoid the non-working area, and construct the boundary of the working area of ​​the lawnmower based on the second target area, the third target area and the subsequent target areas.

Claims

1. A self-learning obstacle avoidance control method for a lawnmower robot, characterized in that, Includes the following steps: S1. Global path planning: The preset processor and positioning module obtain the current location data of the grass, and the preset mobile terminal establishes a communication connection with the processor through the local area network communication module. Based on the location data obtained by the processor, the online map is downloaded from the wide area network. The operator transmits the boundary information of the mowing area to the processor through the mobile terminal, and plans the optimal mowing path through the algorithm. S2. Adaptive obstacle avoidance: After receiving the instructions from the mobile terminal, the lawnmower robot performs the lawnmower task according to the optimal lawnmower path. During the task execution, it judges the shape characteristics of the obstacle based on the lawnmower robot's own motion parameters and performs the lawnmower task along the edge of the obstacle. S3. Real-time update of mowing path: Mark the location of obstacles encountered in the mowing task and update the mowing path in real time. S4. Replenishment path planning: Set the uncut areas along the mowing task as target areas and record the position of the target areas. When the mowing robot performs the mowing task according to the mowing path, it judges the distance between the position of the target area and the updated path and further optimizes the mowing path. The target area setting in step S4 includes the following steps: S41. When the lawnmower robot performs lawnmowing tasks along the mowing path, it scans the grass around the path in real time using a CCD camera, determines the unmowed grass area based on the grass height, and records the coordinates and shape of this area to generate the main area node. S42. Radialize outwards from the main region node, scan the uncut grassland adjacent to the main region node, record its coordinates and shape, and generate sub-region nodes. S43. Establish the relationship between the main region nodes and the sub-region nodes, and generate the target region mesh map based on the distribution of the main region nodes and the sub-region nodes. Calculate the optimal cutting path connecting the main region nodes and multiple sub-region nodes using a genetic algorithm. The step S4 of determining the distance between the target area and the updated path includes the following steps: S401. Obtain the first target area to be processed near the mowing path of the lawn mowing robot, control the lawn mowing robot to move along the mowing path according to the movement command set by the user, and obtain the positioning information of the lawn mowing robot in real time during the movement. Based on the positioning information of the lawn mowing robot, determine the first target area to be processed. S402: Expand the first target area to be processed outward by a preset distance to become the second target area to be processed, and control the lawn mowing robot to move along the second target area to be processed. S403. Obtain environmental information of the lawnmower in the second target area based on the sensors of the lawnmower, determine the boundary of the non-working area in the environmental information, control the lawnmower to move along the boundary of the non-working area to obtain the third target area to avoid the non-working area, and construct the boundary of the working area of ​​the lawnmower based on the second target area, the third target area and the subsequent target areas.

2. The self-learning obstacle avoidance control method for a lawnmower robot according to claim 1, characterized in that: The acquisition of the mowing area boundary information in step S1 includes the following steps: S11. Preset a drone and use the drone's onboard downward-looking binocular passive vision system to traverse the grassland boundary and generate a plot boundary coordinate map; S12. Based on the land parcel boundary coordinate map, identify obstacles within the land parcel boundary and correct the land parcel boundary map; S13. Visual recognition and segmentation technology is used to divide the corrected plot boundaries, determine the drivable area and obstacle boundary information, and plan the optimal drivable area path information.

3. The self-learning obstacle avoidance control method for a lawnmower robot according to claim 2, characterized in that: The algorithm in step S1 is a genetic algorithm.

4. The self-learning obstacle avoidance control method for a lawnmower robot according to claim 3, characterized in that: Determining the shape characteristics of the obstacle in step S2 includes the following steps: S21. A preset CCD camera acquires omnidirectional image features of the obstacle, and a preset processing module divides the obstacle image into n equal regions, setting the color pixels of each small region to be... The number of edge pixels is obtained after edge detection. The processing module combines color pixels and edge pixels into a feature set. S22. Further judgment is made based on the ratio of the number of pixels detected for edge and color. If the ratio is within a set threshold, the area is determined to be an obstacle, as shown in the following formula: ； In the formula, α and β are the upper and lower limits of the ratio between the two; This is the threshold for the number of colors in a region.

5. The self-learning obstacle avoidance control method for a lawnmower robot according to claim 4, characterized in that: The task of mowing along the edge of the obstacle in step S2 includes the following steps: S201. The robot detects obstacles in the grass by using the laser radar and CCD camera preset in the lawnmower, and retrieves the corresponding mowing strategy preset in the lawnmower database through the preset control unit. Then, the lawnmower strategy is converted into instructions and transmitted to the receiving unit of the lawnmower through the preset transmission unit. S202. Control the robot to avoid obstacles and perform grass mowing operations by using a grass-mowing strategy.

6. The self-learning obstacle avoidance control method for a lawnmower robot according to claim 5, characterized in that: The mowing strategy in step S201 involves using the laser radar and CCD camera of the mowing robot to capture and record obstacles in real time, reconstructing the obstacle model through a preset measurement method, and then calculating the optimal mowing path of the mowing robot through a genetic algorithm.

7. The self-learning obstacle avoidance control method for a lawnmower robot according to claim 6, characterized in that: Marking the locations of obstacles encountered in the lawn mowing task in step S3 includes the following steps: S2021. Generate a path map based on the trajectory of the lawnmower robot around the obstacles, and generate corresponding obstacle labels, and store the labels in a preset memory. S2022, The control unit retrieves obstacle tags from the memory and establishes a connection between the tag attributes and the mowing strategy, thus establishing an association.

Images