Patrol robot and automatic obstacle stopping method

Abstract

The invention provides a patrol robot and an autonomous obstacle stopping method, and belongs to the technical field of non-electrical variable control, and the method specifically comprises the steps: subscribing radar point cloud, loading a global point cloud map, calculating a radar-map coordinate transformation matrix, cutting the radar point cloud, transforming the radar point cloud into a map coordinate system, initializing a local cost map, and carrying out the autonomous obstacle stopping of the robot on the basis of the size of the local cost map. Cutting radar point clouds under the global point cloud map and then traversing, establishing a two-dimensional grid cost map, mapping the radar point clouds converted to the map coordinate system to the two-dimensional grid cost map, distinguishing the types of obstacle point clouds to obtain external obstacle point clouds, and determining the number of the external obstacle point clouds and the minimum distance as conditions. Therefore, the reliability of obstacle stopping processing is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of non - electrical variable control, and particularly relates to a patrol robot and an autonomous obstacle - stopping method. Background Art

[0002] In the scenario of narrow underground passages, in order to avoid impact damage to the patrol robot, it is necessary to enable the patrol robot to identify moving obstacles on the road surface and stop autonomously, so as to improve the reliability of the patrol process of the patrol robot.

[0003] Specifically, in the invention patent application CN202310635112.3 "An Intelligent Obstacle - Avoiding Method, System and Medium for a Patrol Robot", the obstacle information is judged by fusing multi - source information data, and the optimal patrol path is planned according to the obstacle information, realizing the intelligent obstacle - avoiding of the patrol robot. However, through analysis, it is not difficult to find the following technical problems:

[0004] Currently, the most commonly used map form is the point - cloud map. However, it is difficult for the point - cloud map to directly reflect the obstacle information and needs to be converted into a two - dimensional grid cost map. Moreover, for the robot to stop obstacles, only the local environment information around is needed. If a global two - dimensional grid map is used, there is information redundancy, consuming more memory resources;

[0005] Conventional obstacle detection methods based on point - cloud information data can usually only detect obstacles, and it is difficult to distinguish between structural static obstacles such as buildings and walls and external obstacles such as pedestrians. Moreover, sometimes there are problems such as false detection and inaccuracy in the obstacle point - cloud data;

[0006] Existing obstacle - stopping methods only calculate the distance between the obstacle and the robot to feedback the instruction to stop moving. If the distance calculation is incorrect, it is easy to cause collisions and the reliability is not strong.

[0007] In summary, the disadvantages of the existing technology are summarized as: failure to reasonably extract the obstacle information in the point - cloud map and convert it into a local two - dimensional grid cost map for description, lack of distinction between static obstacles and external obstacles, and the method of using a single constraint to feedback the stop instruction lacks reliability.

[0008] In view of the above - mentioned technical problems, the present invention provides a patrol robot and an autonomous obstacle - stopping method. Summary of the Invention

[0009] To achieve the object of the present invention, the present invention adopts the following technical solutions:

[0010] According to one aspect of the present invention, an autonomous obstacle - stopping method is provided.

[0011] An autonomous obstacle - stopping method specifically includes:

[0012] S1 Subscribe to the radar point cloud, load the global point cloud map, calculate the radar-map coordinate transformation matrix, and crop and transform the radar point cloud to the map coordinate system;

[0013] S2 Initialize the local cost map. Based on the size of the local cost map, crop the radar point cloud under the global point cloud map and traverse it to establish a two-dimensional grid cost map;

[0014] S3 Map the radar point cloud transformed to the map coordinate system to the two-dimensional grid cost map, and distinguish the obstacle point cloud types to obtain the external obstacle point cloud;

[0015] S4 Determine whether to trigger a stop obstacle signal based on the number and minimum distance of the external obstacle point cloud.

[0016] A further technical solution is that the cropping of the radar point cloud is centered on the centroid of the patrol robot.

[0017] A further technical solution is that cropping and transforming the radar point cloud to the map coordinate system specifically includes:

[0018] Subscribe to the real-time radar point cloud collected by the lidar of the patrol robot and load it into the global point cloud map;

[0019] Subscribe to the positioning information of the patrol robot and use the positioning information to construct the transformation matrix from the radar coordinate system to the map coordinate system;

[0020] Centered on the centroid of the patrol robot, crop and retain the points within the cube range of the preset volume around as the detection range to obtain the cropped radar point cloud, and use the transformation matrix and the cropped radar point cloud to transform the radar point cloud to the map coordinate system.

[0021] A further technical solution is that when cropping the radar point cloud, it is also necessary to crop the radar point cloud within the body area of the patrol robot.

[0022] A further technical solution is that determining whether to trigger a stop obstacle signal specifically includes:

[0023] Accumulatively count the external obstacle point cloud obtained by processing the radar point cloud of each frame to obtain the number of external obstacle point cloud of each frame, and at the same time calculate the closest distance between the external obstacle point cloud of each frame and the center of the robot;

[0024] Set the external obstacle point cloud threshold. If the number of external obstacle point cloud is greater than the external obstacle point cloud threshold or the closest distance is less than the preset distance threshold, then trigger a stop obstacle signal to make the robot stop moving.

[0025] On the other hand, the present application provides a patrol robot that adopts the above-mentioned autonomous obstacle stopping method, specifically including:

[0026] A coordinate transformation module, a cost map construction module, a point cloud type classification module, and an obstacle stopping signal triggering module;

[0027] Among them, the coordinate transformation module is responsible for subscribing to the lidar point cloud, loading the global point cloud map, calculating the radar-map coordinate transformation matrix, and cropping and transforming the lidar point cloud into the map coordinate system;

[0028] The cost map construction module is responsible for initializing the local cost map, cropping the lidar point cloud under the global point cloud map based on the size of the local cost map, and traversing it to establish a two-dimensional grid cost map;

[0029] The point cloud type classification module is responsible for mapping the lidar point cloud transformed into the map coordinate system to the two-dimensional grid cost map, and distinguishing the obstacle point cloud type to obtain the external obstacle point cloud;

[0030] The obstacle stopping signal triggering module is responsible for determining whether to trigger the obstacle stopping signal based on the number and minimum distance of the external obstacle point cloud.

[0031] The beneficial effects of the present invention are as follows:

[0032] The present invention relates to an autonomous obstacle stopping method for a patrol robot. This method requires the lidar carried by the robot to detect the surrounding environment in real time to obtain point cloud data and the already built global point cloud map, subscribe to the pose information of the robot in global positioning, construct a local two-dimensional grid cost map based on the local map point cloud information, process the lidar point cloud, detect static obstacle point clouds and external obstacle point clouds, and finally combine the number and minimum distance information of the external obstacle point cloud, set appropriate thresholds, and timely issue reliable obstacle stopping signals.

[0033] The autonomous obstacle stopping method for the patrol robot proposed by the present invention aims to realize the function of the robot to identify external obstacles and autonomously stop, improve its autonomous movement ability and the reliability of safe travel. When the robot moves along the planned trajectory, it can distinguish static obstacles and external obstacles through the local map, and focus on the number and distance of the external obstacle point cloud, with the characteristics of high real-time performance, simplicity and easy implementation, and good effect.

[0034] Other features and advantages will be described in the subsequent specification. The objectives and other advantages of the present invention are achieved and obtained by the structures specifically pointed out in the specification and the drawings.

[0035] To make the above objectives, features, and advantages of the present invention more obvious and understandable, the following specific preferred embodiments are given, and in conjunction with the accompanying drawings, the detailed description is as follows. Brief Description of the Drawings

[0036] By referring to the drawings and describing in detail its exemplary embodiments, the above and other features and advantages of the present invention will become more apparent.

[0037] Figure 1 is a flowchart of an autonomous obstacle avoidance method;

[0038] Figure 2 is a flowchart of a method for cropping and transforming radar point clouds into the map coordinate system;

[0039] Figure 3 is a flowchart of the specific steps for constructing a two-dimensional grid cost map;

[0040] Figure 4 is a framework diagram of a patrol robot. Detailed Embodiments

[0041] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments of this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this specification.

[0042] Embodiment 1

[0043] To solve the above problems, according to one aspect of the present invention, as Figure 1 shown, an autonomous obstacle avoidance method is provided, which specifically includes:

[0044] S1 Subscribe to the radar point cloud, load the global point cloud map, calculate the radar-map coordinate transformation matrix, and crop and transform the radar point cloud into the map coordinate system;

[0045] Furthermore, the cropping of the radar point cloud is centered on the centroid of the patrol robot.

[0046] It should be noted that, as Figure 2 shown, cropping and transforming the radar point cloud into the map coordinate system specifically includes:

[0047] Subscribe to the real-time radar point cloud collected by the lidar of the patrol robot and load it into the global point cloud map;

[0048] Subscribe to the positioning information of the patrol robot and use the positioning information to construct the transformation matrix from the radar coordinate system to the map coordinate system;

[0049] Taking the centroid of the patrol robot as the center, crop and retain the points within the cube of a preset volume around it as the detection range to obtain the cropped radar point cloud, and transform the radar point cloud to the map coordinate system by using the transformation matrix and the cropped radar point cloud.

[0050] The input radar point cloud data has a frequency of 10HZ. Each time a frame of point cloud is input, it is processed. The cropping of the radar point cloud is centered on the centroid of the robot, and the points within the cube of 4m * 2m * 0.6m around it are cropped and retained as the detection range. At the same time, the standing size of the robot body is 1m * 0.8m * 0.6m. The points that appear within this range are body points and should be filtered out.

[0051] Under the NDT positioning algorithm, the robot coordinate system changes in real time, and the coordinate transformation between the robot coordinate system and the map coordinate system can be monitored. Based on the point cloud in the radar coordinate system, it is transformed to the map coordinate system through coordinate transformation (in this example, it is default that the radar coordinate system is the robot coordinate system). This process can be represented by the following formula:

[0052]

[0053] Among them, represents the transformation matrix from the radar coordinate system to the map coordinate system, represents the point cloud coordinates collected in the radar coordinate system, represents the coordinates of the point cloud after being transformed to the map coordinate system.

[0054] S2 Initialize the local cost map. Based on the size of the local cost map, crop the radar point cloud in the global point cloud map and traverse it to establish a two-dimensional grid cost map;

[0055] Specifically, as Figure 3 shown, the specific steps for constructing the two-dimensional grid cost map are as follows:

[0056] Initialize the local cost map, set the length, width, and resolution of the local cost map, and obtain a grid map based on the length, width, and resolution of the local cost map;

[0057] Taking the centroid of the patrol robot as the center, crop the radar point cloud in the global point cloud map according to the size of the local cost map, traverse the radar point cloud, and calculate the grid index value for the radar point cloud. Set the grid where the radar point cloud is located to the occupied state, and set other grids to the idle state;

[0058] Establish a two-dimensional grid cost map according to the occupied state and idle state of the grid.

[0059] Further, when the index value of the grid is within a preset range, it is determined that the grid is in an occupied state. When the index value of the grid is not within the preset range, it is determined that the grid is in an idle state.

[0060] It should be noted that the resolution and size of the local two-dimensional grid cost map are set. The resolution can be appropriately adjusted according to the map size. In this example, the value of resolution is 0.2m. The size of the local map can also be adjusted considering the actual application scenario. In this example, the length Height = 6m and the width Width = 6m, so as to obtain a grid map with a size of Height / resolution * Width / resolution.

[0061] Crop the map point cloud within a range of 6m * 6m * 0.6m centered on the centroid of the robot. Then align the center of the robot with the center of the grid map, traverse each map point, and calculate the index value of the point projected onto the local grid map:

[0062] IndexH = (point.y - robot.y) / resolution + (Height / 2 * resolution)

[0063] IndexW = (point.x - robot.x) / resolution + (Height / 2 * resolution)

[0064] That is, the point is mapped into the grid map[IndexH][IndexW]. Then assign the value 100 to the grid map[IndexH][IndexW] to represent the occupied state, and assign the value 0 to other grids to represent the idle state.

[0065] S3 Map the radar point cloud transformed into the map coordinate system to the two-dimensional grid cost map, and distinguish the types of obstacle point clouds to obtain the external obstacle point clouds.

[0066] In addition, it should be noted that when cropping the radar point cloud, it is also necessary to crop the radar point cloud within the body area of the patrol robot.

[0067] Specifically, the types of obstacle point clouds include static obstacle point clouds and foreign obstacle point clouds.

[0068] It can be understood that the specific steps for distinguishing the types of obstacle point clouds are as follows:

[0069] Traverse the radar point cloud, and by calculating the index position of the radar point cloud in the map coordinate system relative to the center of the robot, obtain the grid state of each radar point cloud mapped to the two-dimensional grid cost map.

[0070] If the grid status is an occupied state, then the obstacle point cloud type of the radar point cloud corresponding to the grid is included in the static obstacle point cloud. If the grid status is an idle state, then the neighboring grids of the grid are traversed;

[0071] Determine whether the occupancy of the grid status of the neighboring grid meets the requirements. If so, then the obstacle point cloud type of the radar point cloud corresponding to the grid is included in the static obstacle point cloud. Otherwise, the obstacle point cloud type of the radar point cloud corresponding to the grid is included as an external obstacle point cloud.

[0072] Furthermore, the neighboring grids of the grid are the grids within a preset area near the grid.

[0073] It should be noted that determining whether the occupancy of the grid status of the neighboring grid meets the requirements specifically includes:

[0074] Divide the neighboring grids into multiple regions according to the positions of the neighboring grids. Based on the grid status of the neighboring grids in different regions, determine the number of neighboring grids in the occupied state in different regions, and combine the interval grid data between the neighboring grids in the occupied state in different regions to determine the occupancy state values of different regions;

[0075] Take the neighboring grids in the occupied state as occupied grids, and determine the basic occupancy state value of the neighboring grids through the number of the occupied grids and the interval grid data between different occupied grids and the grid;

[0076] Based on the basic occupancy state value of the neighboring grids and the occupancy state value of the region, determine the grid status evaluation quantity of the neighboring grids, and use the grid status evaluation quantity to determine whether the occupancy of the grid status of the neighboring grids meets the requirements.

[0077] Furthermore, using the grid status evaluation quantity to determine whether the occupancy of the grid status of the neighboring grid meets the requirements specifically includes:

[0078] When the grid status evaluation quantity is within the preset state value space, then determine that the occupancy of the grid status of the neighboring grid meets the requirements.

[0079] It can be understood that determining whether the occupancy of the grid status of the neighboring grid meets the requirements specifically includes:

[0080] S21 Take the neighboring grids in the occupied state as occupied grids, and determine whether the number of the occupied grids is greater than the preset number of grids. If so, then determine that the occupancy of the grid status of the neighboring grid meets the requirements. If not, then proceed to the next step;

[0081] S22 Determine whether the number of occupied grids is within a preset grid number range. If so, proceed to the next step. If not, proceed to step

[0082] S23 Based on the interval grid data between different occupied grids and the grid, determine the average value of the number of interval grids between different occupied grids and the grid. Determine whether the average value of the number of interval grids between different occupied grids and the grid is greater than the preset interval grid number. If so, proceed to the next step. If not, determine that the occupancy situation of the grid state of the neighboring grid does not meet the requirements;

[0083] S24 Divide the neighboring grids into multiple regions according to the positions of the neighboring grids. Based on the grid states of the neighboring grids in different regions, determine the number of neighboring grids in different regions that are in the occupied state. Determine whether there is a region where the number of neighboring grids in the occupied state is greater than the preset number threshold. If so, proceed to the next step. If not, determine that the occupancy situation of the grid state of the neighboring grid does not meet the requirements;

[0084] S25 Determine the occupancy state values of different regions through the number of neighboring grids in the occupied state in different regions and the interval grid data between the neighboring grids in the occupied state in different regions. Determine the basic occupancy state value of the neighboring grid through the number of occupied grids and the interval grid data between different occupied grids and the grid;

[0085] S26 Determine the grid state evaluation quantity of the neighboring grid based on the basic occupancy state value of the neighboring grid and the occupancy state value of the region, and use the grid state evaluation quantity to determine whether the occupancy situation of the grid state of the neighboring grid meets the requirements.

[0086] It should be noted that after the radar point cloud is converted to the map coordinate system, traverse the radar point cloud and process the process of mapping the radar point cloud to the grid map in the same way. If the grid value mapped to is 100, it is in the occupied state, then this radar point is identified as a static obstacle point. If the grid value mapped to is 0, continue to traverse the neighboring grids of this grid. If there are occupied grids among the neighboring grids, then this point cloud is also identified as a static obstacle point. If there are no occupied grids among the neighboring grids, then this point is identified as an external obstacle point.

[0087] Thus, static obstacle points or external obstacle points can be distinguished. Static obstacles have less impact during the normal patrol of the robot, while external obstacles are often some dynamic obstacles and are more likely to have an adverse impact on the autonomous movement of the robot.

[0088] S4 determines whether to trigger an obstacle stopping signal based on the number and minimum distance of the foreign obstacle point cloud.

[0089] Furthermore, determining whether to trigger an obstacle stopping signal specifically includes:

[0090] Accumulatively count the foreign obstacle point clouds obtained by processing the radar point clouds of each frame to obtain the number of foreign obstacle point clouds per frame, and at the same time calculate the closest distance between the foreign obstacle point clouds of each frame and the center of the robot;

[0091] Set a foreign obstacle point cloud threshold. If the number of foreign obstacle point clouds is greater than the foreign obstacle point cloud threshold or the closest distance is less than the preset distance threshold, then trigger an obstacle stopping signal to make the robot stop moving.

[0092] After traversing the radar point clouds, count from the foreign obstacle point clouds, and at the same time calculate the minimum value of the distances of these points from the center of the robot. This process is shown as follows:

[0093] Mindistance=min{(P x -R x ) 2 +(R y -R y ) 2 +(P z -R z ) 2}

[0094] Among them, [P x ,P y ,P z represents the xyz three-dimensional coordinates of the foreign obstacle point set, and [R x ,R y ,R z represents the xyz three-dimensional coordinates of the foreign obstacle point set.

[0095] The number of foreign obstacle point clouds TH num and the minimum distance threshold TH dis of this method can both be numerically adjusted for different environments. In this example, TH num is set to 10, and TH dis is set to 1m. If the number of point clouds exceeds 20 or the closest distance is less than 1m, then trigger an obstacle stopping signal for the robot. When the number of radar lines is higher and the point clouds are denser, consider increasing the value of TH num . If the ground area is large and the environment is open, consider increasing the value of TH dis .

[0096] Example 2

[0097] On the other hand, as Figure 4As shown in the figure, the present application provides a patrol robot, which adopts the above-mentioned autonomous obstacle stopping method, specifically including:

[0098] A coordinate transformation module, a cost map construction module, a point cloud type classification module, and an obstacle stopping signal triggering module;

[0099] Among them, the coordinate transformation module is responsible for subscribing to the lidar point cloud, loading the global point cloud map, calculating the radar-map coordinate transformation matrix, and cropping and transforming the lidar point cloud to the map coordinate system;

[0100] The cost map construction module is responsible for initializing the local cost map, based on the size of the local cost map, cropping the lidar point cloud under the global point cloud map and traversing it to establish a two-dimensional grid cost map;

[0101] The point cloud type classification module is responsible for mapping the lidar point cloud transformed to the map coordinate system to the two-dimensional grid cost map, distinguishing the obstacle point cloud types to obtain the external obstacle point cloud;

[0102] The obstacle stopping signal triggering module is responsible for determining whether to trigger the obstacle stopping signal based on the number and minimum distance of the external obstacle point cloud.

[0103] Based on the above embodiments, the present invention has the following beneficial effects:

[0104] The present invention relates to an autonomous obstacle stopping method for a patrol robot. This method requires the lidar carried by the robot body to detect the surrounding environment in real time to obtain point cloud data, and the already built global point cloud map, subscribe to the pose information of the robot in global positioning, construct a local two-dimensional grid cost map based on the local map point cloud information, process the lidar point cloud, detect static obstacle point clouds and external obstacle point clouds, and finally combine the number and minimum distance information of the external obstacle point cloud, set appropriate thresholds, and timely issue reliable obstacle stopping signals.

[0105] The autonomous obstacle stopping method for the patrol robot proposed by the present invention aims to realize the function of the robot to identify external obstacles and autonomously stop, improve its autonomous movement ability and the reliability of safe travel. When the robot moves along the planned trajectory, it can distinguish static obstacles and external obstacles through the local map, and focus on the number and distance of the external obstacle point cloud, with the characteristics of high real-time performance, simplicity and good effect.

[0106] An example of the present invention is deployed using a Velodyne VLP16 lidar, which publishes a frame of point cloud data every 0.1 s. The global point cloud map is a map in pcd format. The point cloud map has been filtered to remove redundant points using the voxel filtering method, and the initial pose at the start of mapping is used as the origin coordinate system of the map. The robot selected is a bionic quadruped robot, and the software environment is the ROS Melodic robot operating system based on Ubuntu 18.04.

[0107] The robot positioning information required by the present invention can be obtained by using the NDT algorithm for global map relocalization, and the three-dimensional pose of the robot in the map can be output in real time.

[0108] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the differences between each embodiment and other embodiments are emphasized. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments.

[0109] The above describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain implementations, multitasking and parallel processing are also possible or may be advantageous.

[0110] The above is only one or more embodiments of this specification and is not used to limit this specification. For those skilled in the art, one or more embodiments of this specification can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of one or more embodiments of this specification shall be included within the scope of the claims of this specification.

Claims

1. An autonomous obstacle prevention method, characterized in that: Specifically include: Subscribe to the radar point cloud, load the global point cloud map, calculate the radar-map coordinate transformation matrix, crop the radar point cloud, and transform it to the map coordinate system; Initialize a local cost map, and based on the size of the local cost map, crop the radar point cloud under the global point cloud map and traverse it to establish a two-dimensional grid cost map; Mapping the radar point cloud transformed into the map coordinate system to a two-dimensional grid cost map, distinguishing obstacle point cloud types to obtain external obstacle point clouds; Whether to trigger the obstacle stop signal is determined based on the number and minimum distance of the external obstacle point cloud.

2. The autonomous obstacle stopping method according to claim 1, characterized in that: The radar point cloud is cropped with the center of mass of the patrol robot as the center.

3. The autonomous obstacle stopping method according to claim 1, characterized in that: Crop the radar point cloud and transform it to the map coordinate system, including: Subscribe to the real-time radar point cloud collected by the patrol robot's lidar and load it into the global point cloud map; Subscribe to the positioning information of the patrol robot, and use the positioning information to construct a transformation matrix from the radar coordinate system to the map coordinate system; Taking the center of mass of the patrol robot as the center, points within a cube range of a preset volume are cropped and retained as the detection range to obtain a cropped radar point cloud, and the radar point cloud is transformed into a map coordinate system using the transformation matrix and the cropped radar point cloud.

4. The autonomous obstacle stopping method according to claim 3, characterized in that: When the radar point cloud is clipped, the radar point cloud within the body area of ​​the patrol robot also needs to be clipped.

5. The autonomous obstacle stopping method according to claim 1, characterized in that: The specific steps of constructing the two-dimensional grid cost map are: Initialize the local cost map, set the length, width and resolution of the local cost map, and obtain a grid map with the length, width and resolution of the local cost map; Taking the center of mass of the patrol robot as the center, the radar point cloud under the global point cloud map is clipped according to the size of the local cost map, the radar point cloud is traversed, and the index value of the grid displayed by the radar point cloud is used to set the grid where the radar point cloud is located to an occupied state, and other grids to an idle state; A two-dimensional grid cost map is established according to the occupied state and the idle state of the grid.

6. The autonomous obstacle stopping method according to claim 5, characterized in that: When the index value of the grid is within a preset range, it is determined that the grid is set to an occupied state, and when the index value of the grid is not within the preset range, it is determined that the grid is in an idle state.

7. The autonomous obstacle stopping method according to claim 1, characterized in that: The specific steps of distinguishing the obstacle point cloud types are: Traversing the radar point cloud, obtaining the grid state of each radar point cloud mapped to the two-dimensional grid cost map by calculating the index position of the radar point cloud in the map coordinate system relative to the center of the robot; If the grid state is occupied, the obstacle point cloud type of the radar point cloud corresponding to the grid is counted into the static obstacle point cloud; if the grid state is idle, the neighboring grids of the grid are traversed; It is determined whether the grid state occupancy of the neighboring grid meets the requirements. If so, the obstacle point cloud type of the radar point cloud corresponding to the grid is counted as a static obstacle point cloud. Otherwise, the obstacle point cloud type of the radar point cloud corresponding to the grid is counted as an external obstacle point cloud.

8. The autonomous obstacle stopping method according to claim 7, characterized in that: The adjacent grids of the grid are grids within a preset area near the grid.

9. The autonomous obstacle stopping method according to claim 1, characterized in that: Determine whether to trigger the stop signal, including: The external obstacle point cloud obtained by processing the radar point cloud of each frame is accumulated and counted to obtain the number of external obstacle point clouds of each frame, and the shortest distance between the external obstacle point cloud of each frame and the center of the robot is calculated; Set the external obstacle point cloud threshold. If the number of external obstacle point clouds is greater than the external obstacle point cloud threshold or the closest distance is less than the preset distance threshold, the stop signal is triggered to stop the robot from moving.

10. A patrol robot, using an autonomous obstacle avoidance method according to any one of claims 1 to 9, characterized in that: Specifically include: Coordinate transformation module, cost map construction module, point cloud type classification module, obstacle signal triggering module; The coordinate transformation module is responsible for subscribing to the radar point cloud, loading the global point cloud map, calculating the radar-map coordinate transformation matrix, and clipping and transforming the radar point cloud to the map coordinate system; The cost map is responsible for constructing a module to initialize a local cost map, based on the size of the local cost map, and then traverses the radar point cloud under the global point cloud map to establish a two-dimensional grid cost map; The point cloud type classification module is responsible for mapping the radar point cloud transformed into the map coordinate system to the two-dimensional grid cost map, distinguishing the obstacle point cloud type to obtain the external obstacle point cloud; The obstacle stop signal triggering module is responsible for determining whether to trigger the obstacle stop signal based on the number of the external obstacle point cloud and the minimum distance.

Images