2D laser point cloud channel robot road condition recognition method and system and storage medium

Abstract

The invention belongs to the technical field of laser point cloud feature extraction, and particularly relates to a 2D laser point cloud channel robot road condition recognition method and system and a storage medium. Comprising the steps that S1, standard point cloud groups of various channel types are generated, and the channel types comprise straight roads, L-shaped intersections, T-shaped intersections and crossroads; s2, acquiring a single-frame actual 2D laser point cloud of the current position of the robot; s3, performing nonlinear correlation analysis on the single-frame actual 2D laser point cloud and the standard point cloud group, identifying the channel type of the robot, and estimating the relative pose of the robot; s4, based on the estimated pose, performing fine registration on the single-frame actual 2D laser point cloud and the standard point cloud of the corresponding channel type by using a point cloud registration method; and S5, carrying out clustering segmentation on the point cloud after fine registration, and extracting channel wall surface features, angular point features and obstacle features. The invention provides various road condition standard point cloud generation methods considering the pose of the robot.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of laser point cloud feature extraction, and specifically relates to a 2D laser point cloud channel robot road condition recognition method, system and storage medium. BACKGROUND

[0002] The perception of the surrounding environment (including the channel wall surface and other obstacles) of a channel robot is the basis for its positioning and navigation. At present, the main ways for robots to perceive the surrounding environment are ultrasonic ranging perception, visual perception, laser radar perception, etc. The ultrasonic ranging sensor is arranged on both sides of the chassis of the robot, which can obtain the relative position of the robot and the channel wall surface. For example, Chinese Invention Patent CN201310390231.3, entitled "Ultrasonic navigation device applied to autonomous mobile vehicle in greenhouse", although the patent can perceive the relative pose of the robot and the straight wall, the robot cannot cope with the problem of recognizing various intersections. The camera can perceive the objects within the field of view of the robot camera through target recognition algorithm, but it cannot perceive the environment outside its field of view. The laser radar can perform 360-degree non-blind angle perception around the robot. For example, Chinese Invention Patent CN202211309903.9, entitled "Autonomous navigation method for robot driving in channel based on single-line laser radar", the patent can cluster and segment the wall point cloud of various intersections, but the robot cannot cope with the problem of simultaneously recognizing obstacles in the channel.

[0003] In summary, the existing technology has the following problems: (1) The ultrasonic ranging device cannot perceive the intersection conditions of the robot.

[0004] (2) The visual perception range of the camera is limited.

[0005] (3) The current clustering recognition algorithm of single-line laser radar cannot simultaneously recognize the channel wall surface and obstacles. SUMMARY

[0006] The present application provides a 2D laser point cloud channel robot road condition recognition method, system and storage medium to solve the above problems.

[0007] The present application adopts the following technical solution: a 2D laser point cloud channel robot road condition recognition method, comprising: S1: generating a standard point cloud group of multiple channel types, the channel types including straight, L-shaped intersection, T-shaped intersection and cross intersection; S2: acquiring a single-frame actual 2D laser point cloud of the current position of the robot; S3: performing nonlinear correlation analysis on the single-frame actual 2D laser point cloud and the standard point cloud group, identifying the channel type where the robot is located, and estimating the relative pose of the robot; S4: Based on the estimated pose, the point cloud registration method is used to finely register the actual 2D laser point cloud of a single frame with the standard point cloud of the corresponding channel type. S5: Cluster and segment the finely registered point cloud to extract channel wall features, corner features, and obstacle features.

[0008] In some embodiments, S1 includes: S11: For different channel types, the wall laser area is divided into multiple intervals, and the laser line length of the point cloud in each interval is calculated based on the relative pose relationship between the robot and the channel. S12: Transform the laser line angle and laser line length in the polar coordinate system to the rectangular coordinate system to finally obtain the expression of the standard point cloud of the wall.

[0009] In some embodiments, S11 includes: For a straight path, the wall laser area is divided into 4 intervals. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For L-shaped intersections, the wall laser area is divided into 6 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For T-junctions, the wall laser area is divided into 6 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For an intersection, the wall laser area is divided into 8 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: in L This is the length of the laser line; The radius of the laser range of interest; the channel width is w; the relative distance between the lidar and the channel, with a lateral offset and b longitudinal offset; The laser angle is the angle between the laser line and the negative X-axis of the lidar coordinate system.

[0010] In some embodiments, S12 includes: According to the robot's azimuth angle Rotate the point cloud in the Cartesian coordinate system Degrees, record the point cloud number c that is closest to the laser angle in the negative X-axis direction of the lidar coordinate system; The laser line length sequence that is 0 is The data is numbered from c to the last. The first part of the laser line length sequence, Data composition numbered from 1 to c-1 The second part: ; in This indicates the given lateral offset a, longitudinal offset b, and azimuth angle of the lidar. The laser line length sequence of the standard point cloud in a k-type intersection, where S represents a straight road, L represents an L-type intersection, T represents a T-type intersection, and C represents a crossroads.

[0011] In some embodiments, S3 includes: The correlation coefficient between the actual 2D laser point cloud and the standard point cloud in a single frame was calculated using Kendall's nonlinear correlation analysis method. in For Kendall correlation analysis operators, This represents a sequence consisting of the lengths of laser points in a single frame of actual 2D laser point cloud. Obtain the most relevant data to the actual 2D laser point cloud in a single frame. ,according to The subscript can identify the type of channel the robot is in and roughly estimate the robot's pose.

[0012] In some embodiments, step S4 includes: An iterative nearest-point algorithm is used to align the actual 2D laser point cloud of a single frame with the standard channel point cloud of the current road condition. The aligned single-frame actual 2D laser point cloud is rotated to a position where the robot's heading angle is 0, making the point cloud uniform, so that the wall point cloud is distributed horizontally or vertically.

[0013] In some embodiments, S5 includes: S51: Point cloud extraction and segmentation of wall regions: including point clouds of horizontal wall regions, vertical wall regions, and corner regions; S52: Clustering and segmentation of point clouds of obstacles outside the wall: Point clouds falling outside the wall area are considered obstacles, and different obstacles are identified by a clustering algorithm based on their distribution density.

[0014] In some embodiments, in S51, different road surface conditions are classified as follows: For L-shaped intersections and crossroads, the walls consist of corner-shaped surfaces; for straight roads, the walls are without corner-shaped surfaces; for T-shaped intersections, the walls include both corner-shaped and corner-shaped surfaces. The wall surface is divided into two types: wall surface area with corners and wall surface area without corners. 1) A wall region with corners is represented as: in For the wall region at the i-th corner point, This is a reference point on the standard wall surface; here, the reference point is a corner point. This represents the offset of the point cloud relative to the standard wall surface, with the specific value determined by the actual sensor noise level and wall surface unevenness. For horizontal wall area, For vertical wall areas and For the corner region, horizontal means parallel to the x-axis, and vertical means perpendicular to the x-axis. The direction of the wall line points from the corner side to the other side. When the horizontal wall line direction is in the same direction as the positive x-axis, h is 1, otherwise it is -1; when the vertical wall line direction is in the same direction as the positive y-axis, v is -1, otherwise it is 1.

[0015] 2) A wall region without corner points is represented as: The i-th horizontal wall region without corner points is: The i-th vertical wall region without corner points is: The reference point here It is any point on the standard wall surface.

[0016] A road condition recognition system for a channel robot includes: a processor; a memory storing a computer program; and a lidar sensor for acquiring a single-frame 2D lidar point cloud. When the processor executes the computer program, it implements a single-frame 2D lidar point cloud road condition recognition method for the channel robot.

[0017] A computer-readable storage medium storing a computer program that, when executed by a processor, implements a single-frame 2D laser point cloud channel robot road condition recognition method.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention provides a method for generating standard point clouds for various road conditions that take into account robot pose, and provides a correlation analysis database for robot road condition recognition.

[0019] 2. This invention establishes a correlation analysis and identification model for various road conditions in the channel based on a single-frame 2D laser point cloud, and performs coarse registration between the actual point cloud and the standard point cloud.

[0020] 3. This invention further refines the registration of coarsely registered point clouds based on road condition recognition and proposes a wall region clustering and segmentation method that can simultaneously extract point clouds containing wall features, corner features, and obstacle features. Attached Figure Description

[0021] Figure 1 This describes the working scenario of a channel robot. Figure 2For the generation of standard point clouds for straight sections; Figure 3 Generate a standard point cloud for an L-shaped intersection; Figure 4 Generate a standard point cloud for a T-shaped intersection; Figure 5 Generate a standard point cloud for a crossroads; Figure 6 To obtain a standard point cloud that takes into account the robot's azimuth angle; Figure 7 A detailed schematic diagram of robot pose under various typical road conditions (taking a single-direction position divided into 3 equal parts and an angle divided into 8 equal parts as an example). Figure 8 For point cloud registration and clustering segmentation; Figure 9 For simulation environment; Figure 10 The correlation coefficient; Figure 11 For straight road condition identification and feature extraction; Figure 12 For L-shaped intersection traffic condition recognition and feature extraction; Figure 13 For traffic condition recognition and feature extraction at T-junctions; Figure 14 For traffic condition recognition and feature extraction at intersections. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Typical road conditions include straight roads, L-shaped intersections, T-shaped intersections, and crossroads, such as... Figure 1 As shown, different road conditions have their own unique point cloud distribution patterns. To achieve autonomous driving of the robot, it is necessary to accurately identify the road conditions of the passage where the robot is located without mapping, which forms the basis of autonomous navigation. To this end, we first establish the nonlinear representation of the single-frame 2D point cloud of various passages under different robot poses, and generate standard point clouds for various passages based on this. Then, we analyze the nonlinear correlation between the actual point cloud of a single frame and the standard point cloud under different poses of different passage types to accurately identify the current passage type of the robot and roughly estimate the robot's pose relative to the wall. Through fine registration and clustering segmentation methods, we extract the features of the passage wall, corner features, and obstacle features.

[0024] Specifically, a 2D laser point cloud-based road condition recognition method for a channel robot includes: S1: Generate standard point cloud groups with multiple channel types, including straight roads, L-shaped intersections, T-shaped intersections, and crossroads; S2: Obtain the actual 2D laser point cloud of the robot's current position in a single frame; S3: Perform nonlinear correlation analysis on the single-frame actual 2D laser point cloud and the standard point cloud group to identify the channel type where the robot is located and estimate the robot's relative pose. S4: Based on the estimated pose, the point cloud registration method is used to finely register the actual 2D laser point cloud of a single frame with the standard point cloud of the corresponding channel type. S5: Cluster and segment the finely registered point cloud to extract channel wall features, corner features, and obstacle features.

[0025] In a specific embodiment, S1 includes: S11: For different channel types, the wall laser area is divided into multiple intervals, and the laser line length of the point cloud in each interval is calculated based on the relative pose relationship between the robot and the channel. S12: Transform the laser line angle and laser line length in the polar coordinate system to the rectangular coordinate system to finally obtain the expression of the standard point cloud of the wall.

[0026] Specifically, the distribution pattern of the wall point cloud and the channel parameters (type and width) (etc.), relative distance of lidar (lateral offset) Vertical offset ) and azimuth and laser angle Related. Since the wall point cloud is generated by a lidar starting from a 0-degree laser angle and increasing counterclockwise at certain laser interval angles, the laser line length is calculated using trigonometric geometry based on the relative pose of the robot and the channel. The laser line angle and length in polar coordinates are then transformed to a rectangular coordinate system, ultimately obtaining the representation of the standard wall point cloud. In a specific embodiment, step S11 includes: For straight roads, such as Figure 2 As shown, the wall laser area is divided into 4 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For L-shaped intersections, such as Figure 3 As shown, the wall laser area is divided into 6 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For T-junctions, such as Figure 4As shown, the wall laser area is divided into 6 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For intersections, such as Figure 5 As shown, the wall laser area is divided into 8 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: in L This is the length of the laser line; The radius of the laser range of interest; the channel width is w; the relative distance between the lidar and the channel, with a lateral offset and b longitudinal offset; The laser angle is the angle between the laser line and the negative X-axis of the lidar coordinate system.

[0027] In a specific embodiment, S12 includes: According to the robot's azimuth angle Rotate the Cartesian coordinate system Degrees, record the point cloud number c that is closest to the laser angle in the negative X-axis direction of the lidar coordinate system, such as Figure 6 As shown; The laser line length sequence that is 0 is The data is numbered from c to the last. The first part of the laser line length sequence, Data composition numbered from 1 to c-1 The second part: ; in This indicates the given lateral offset a, longitudinal offset b, and azimuth angle of the lidar. The sequence of laser line lengths in the standard point cloud at a k-type intersection, where S represents a straight road, L represents an L-shaped intersection, T represents a T-shaped intersection, and C represents a crossroads, as shown below. Figure 7 The diagram shows a breakdown of the robot pose with different parameters.

[0028] In a specific embodiment, S3 includes: The correlation coefficient between the actual 2D laser point cloud and the standard point cloud in a single frame was calculated using Kendall's nonlinear correlation analysis method. in For Kendall correlation analysis operators, This represents a sequence consisting of the lengths of laser points in a single frame of actual 2D laser point cloud. Obtain the most relevant data to the actual 2D laser point cloud in a single frame. ,according to The subscript can identify the type of channel the robot is in and roughly estimate the robot's pose.

[0029] In a specific embodiment, S4 includes: S41: The iterative nearest point algorithm is used to align the actual 2D laser point cloud of a single frame with the standard channel point cloud of the current road condition. S42: Rotate the aligned single-frame actual 2D laser point cloud to a position where the robot's heading angle is 0, making the point cloud uniform, so that the wall point cloud is distributed horizontally or vertically.

[0030] Based on a rough estimate of the robot's pose, a point cloud registration method based on Iterative Closest Point (ICP) is used to align the actual point cloud of a single frame with the standard channel point cloud of the current road condition. Figure 8 As shown. Then, the point cloud is rotated to a position where the robot's heading angle is 0, making the point cloud uniform, so that the wall point cloud is distributed horizontally or vertically, providing preprocessing data for subsequent point cloud regions.

[0031] In a specific embodiment, S5 includes: S51: Point cloud extraction and segmentation of wall regions: including point clouds of horizontal wall regions, vertical wall regions, and corner regions. The wall types vary depending on different road conditions, and are classified as follows: For L-shaped intersections and crossroads, the walls consist of corner-shaped surfaces; for straight roads, the walls are without corner-shaped surfaces; for T-shaped intersections, the walls include both corner-shaped and corner-shaped surfaces. The wall surface is divided into two types: wall surface area with corners and wall surface area without corners. 1) A wall region with corners is represented as: in For the wall region at the i-th corner point, This is a reference point on the standard wall surface; here, the reference point is a corner point. This represents the offset of the point cloud relative to the standard wall surface, with the specific value determined by the actual sensor noise level and wall surface unevenness. For horizontal wall area, For vertical wall areas and For the corner region, horizontal means parallel to the x-axis, and vertical means perpendicular to the x-axis. The direction of the wall line points from the corner side to the other side. When the horizontal wall line direction is in the same direction as the positive x-axis, h is 1, otherwise it is -1; when the vertical wall line direction is in the same direction as the positive y-axis, v is -1, otherwise it is 1.

[0032] 2) A wall region without corner points is represented as: The i-th horizontal wall region without corner points is: The i-th vertical wall region without corner points is: The reference point here It is any point on the standard wall surface.

[0033] Based on the road condition recognition results, the wall regions corresponding to the road conditions are represented, and the point clouds falling in different wall regions are extracted, thereby achieving clustering and segmentation of the point clouds of the wall regions.

[0034] S52: Clustering and segmentation of point clouds of obstacles outside the wall: Point clouds falling outside the wall area are considered obstacles, and different obstacles are identified by a clustering algorithm based on their distribution density.

[0035] Simulation verification Create a Gazebo instance as follows Figure 9 The simulation environments for the four road conditions shown in this paper use the proposed method for road condition identification and clustering segmentation.

[0036] This paper uses an L-shaped intersection as an example to illustrate the Kendall nonlinear correlation analysis method for identifying road conditions. Kendall analysis is performed on the actual point cloud of the L-shaped intersection and standard point cloud groups (3 equal divisions for position and 36 equal divisions for angle) of straight roads, L-shaped intersections, T-shaped intersections, and crossroads. The maximum correlation values ​​between this point cloud and the standard point cloud groups of straight roads, L-shaped intersections, T-shaped intersections, and crossroads are 0.3877, 0.7618, 0.5279, and 0.4298, respectively. Since the maximum value of 0.7618 corresponds to the standard point cloud of an L-shaped intersection, the road condition where the robot is located can be identified as an L-shaped intersection. Figure 10 The correlation matrix shown is between the actual point cloud and the standard point cloud group of the L-shaped intersection. It consists of 36 rows (each row represents a different point cloud rotation angle) and 9 columns (each column represents a different position offset). The pose parameters of the standard point cloud with a maximum value of 0.7618 are a rotation angle of 340 degrees, an offset a of 0.5 meters, and an offset b of 0.5 meters. This standard point cloud, as shown in the road condition recognition map, provides the initial pose for ICP fine registration.

[0037] Figure 11 , 12 Figures 13 and 14 show the intermediate processes and results of road condition recognition and clustering segmentation for straight roads, L-shaped intersections, T-shaped intersections, and crossroads, respectively. It can be seen that the wall point clouds, corner point clouds, and obstacle point clouds of each channel are effectively and synchronously identified.

[0038] A road condition recognition system for a channel robot includes: a processor; a memory storing a computer program; and a lidar sensor for acquiring a single-frame 2D lidar point cloud. When the processor executes the computer program, it implements a road condition recognition method for a channel robot based on a single-frame 2D lidar point cloud.

[0039] A feature extraction system for a channel robot includes: a processor; a memory storing a computer program; and a lidar sensor for acquiring a single-frame 2D laser point cloud. When the processor executes the computer program, it implements a clustering and segmentation method for the channel robot walls and obstacles, such as the single-frame 2D laser point cloud.

[0040] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A 2D laser point cloud-based road condition recognition method for a channel robot, characterized in that, include: S1: Generate standard point cloud groups with multiple channel types, including straight roads, L-shaped intersections, T-shaped intersections, and crossroads; S2: Obtain the actual 2D laser point cloud of the robot's current position in a single frame; S3: Perform nonlinear correlation analysis on the single-frame actual 2D laser point cloud and the standard point cloud group to identify the channel type where the robot is located and estimate the robot's relative pose. S4: Based on the estimated pose, the point cloud registration method is used to finely register the actual 2D laser point cloud of a single frame with the standard point cloud of the corresponding channel type. S5: Cluster and segment the finely registered point cloud to extract channel wall features, corner features, and obstacle features.

2. The 2D laser point cloud channel robot road condition recognition method according to claim 1, characterized in that, S1 includes: S11: For different channel types, the wall laser area is divided into multiple intervals, and the laser line length of the point cloud in each interval is calculated based on the relative pose relationship between the robot and the channel. S12: Transform the laser line angle and laser line length in the polar coordinate system to the rectangular coordinate system to finally obtain the expression of the standard point cloud of the wall.

3. The 2D laser point cloud channel robot road condition recognition method according to claim 2, characterized in that, S11 includes: For a straight path, the wall laser area is divided into 4 intervals. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For L-shaped intersections, the wall laser area is divided into 6 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For T-junctions, the wall laser area is divided into 6 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: For an intersection, the wall laser area is divided into 8 sections. The formula for calculating the wall laser area and the corresponding laser line length is as follows: in L The length of the laser line; The radius of the laser range of interest; the channel width is w; the relative distance between the lidar and the channel, with a lateral offset and b longitudinal offset; The laser angle is the angle between the laser line and the negative X-axis of the lidar coordinate system.

4. The 2D laser point cloud channel robot road condition recognition method according to claim 3, characterized in that, S12 includes: According to the robot's azimuth angle Rotate the point cloud in the Cartesian coordinate system Degrees, record the point cloud number c that is closest to the laser angle in the negative X-axis direction of the lidar coordinate system; The laser line length sequence that is 0 is The data is numbered from c to the last. The first part of the laser line length sequence, Data composition numbered from 1 to c-1 The latter part: ; in This indicates the given lateral offset a, longitudinal offset b, and azimuth angle of the lidar. The laser line length sequence of the standard point cloud in a k-type intersection, where S represents a straight road, L represents an L-type intersection, T represents a T-type intersection, and C represents a crossroads.

5. The 2D laser point cloud channel robot road condition recognition method according to claim 1, characterized in that, S3 includes: The correlation coefficient between the actual 2D laser point cloud and the standard point cloud in a single frame was calculated using Kendall's nonlinear correlation analysis method. in For Kendall correlation analysis operators, This represents a sequence of laser point lengths representing a single frame of actual 2D laser point cloud. Obtain the most relevant data to the actual 2D laser point cloud in a single frame. ,according to The subscript can identify the type of channel the robot is in and roughly estimate the robot's pose.

6. The 2D laser point cloud channel robot road condition recognition method according to claim 1, characterized in that, S4 includes: An iterative nearest-point algorithm is used to align the actual 2D laser point cloud of a single frame with the standard channel point cloud of the current road condition. The aligned single-frame actual 2D laser point cloud is rotated to a position where the robot's heading angle is 0, making the point cloud uniform, so that the wall point cloud is distributed horizontally or vertically.

7. The 2D laser point cloud channel robot road condition recognition method according to claim 1, characterized in that, S5 includes: S51: Point cloud extraction and segmentation of wall regions: including point clouds of horizontal wall regions, vertical wall regions, and corner regions; S52: Clustering and segmentation of point clouds of obstacles outside the wall: Point clouds falling outside the wall area are considered obstacles, and different obstacles are identified by a clustering algorithm based on their distribution density.

8. The 2D laser point cloud channel robot road condition recognition method according to claim 1, characterized in that, In S51, different road surface conditions are classified as follows: For L-shaped intersections and crossroads, the walls consist of corner-shaped surfaces; for straight roads, the walls are without corner-shaped surfaces; for T-shaped intersections, the walls include both corner-shaped and corner-shaped surfaces. The wall surface is divided into two types: wall surface area with corners and wall surface area without corners. 1) A wall region with corners is represented as: in For the wall region at the i-th corner point, This is a reference point on the standard wall surface; here, the reference point is a corner point. This represents the offset of the point cloud relative to the standard wall surface, with the specific value determined by the actual sensor noise level and wall surface unevenness. For horizontal wall area, For vertical wall areas and For the corner region, horizontal means parallel to the x-axis, and vertical means perpendicular to the x-axis. The direction of the wall line points from the corner side to the other side. When the horizontal wall line direction is in the same direction as the positive x-axis, h is 1, otherwise it is -1; when the vertical wall line direction is in the same direction as the positive y-axis, v is -1, otherwise it is 1. 2) A cornerless wall region is represented as: The i-th horizontal wall region without corner points is: The i-th vertical wall region without corner points is: The reference point here It is any point on the standard wall surface.

9. A system, characterized in that, include: processor; Memory, which stores computer programs; A lidar sensor is used to acquire a single frame of 2D laser point cloud. When the processor executes the computer program, it implements the 2D laser point cloud channel robot road condition recognition method as described in any one of claims 1-8.

10. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the 2D laser point cloud channel robot road condition recognition method as described in any one of claims 1-8.

Images