Bulk yard mobile robot 2d-slam method and system

Abstract

The application provides a bulk storage yard mobile robot 2D-SLAM method and system, and belongs to the technical field of indoor positioning and map construction. The problems of low map construction precision and safety accidents such as equipment collision in iron ore, coal powder and other environments are solved. The technical scheme is as follows: step one, environment information scanning, the scrap steel grab machine scans the scrap steel yard environment through the far-field laser radar when starting work, and determines the scrap steel pile position and drivable area; step two, global map construction, a global grid map of the scrap steel yard is established by using a map construction method based on the Cartographer framework; step three, obstacle avoidance map generation, when the scrap steel grab machine is working, the near-field laser radar scans the environment information around the grab machine, and a local obstacle avoidance map is generated in real time. The application has the beneficial effects that the application can improve the mapping precision and improve the obstacle avoidance perception ability of the grab machine.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of indoor positioning and map construction, and particularly relates to a 2D-SLAM method and system for mobile robots in a bulk material yard. BACKGROUND

[0002] Manual operation in a harsh environment of bulk materials such as iron ore and coal is difficult, and mobile robots can perform inventory, sampling and other operations. Traditional inventory, sampling and other operations rely on manual operation, which is inefficient and dangerous in a hazardous environment. Therefore, the demand for related automated operations in the industry is increasingly urgent. As a key link for mobile robots to achieve automated operation, simultaneous localization and mapping (SLAM) is particularly important. The scene map constructed by SLAM technology is an important support for the navigation and path planning of a grabber.

[0003] Currently, SLAM technology can be divided into two categories: the first category is camera-based visual SLAM, which requires high environmental lighting and is difficult to distinguish between regions with similar texture features; the second category is laser radar-based SLAM, which can operate in poor light conditions and has stronger universality. 2D-SLAM is a laser SLAM method, which refers to a SLAM method for obtaining a two-dimensional map of the plane in which it is located through a laser sensor in a two-dimensional plane. Due to its high mapping accuracy, stable performance and low price, it is widely used. However, there are still several problems with the application of 2D-SLAM technology in a scrap yard environment:

[0004] (1) The environment of the bulk material yard is complex, and the mobile robot is prone to collisions and other safety accidents during autonomous operation.

[0005] (2) The bulk material piles are irregularly placed, and the features scanned by the laser radar are prone to false matching, resulting in low precision of the constructed map.

[0006] How to solve the above technical problems is the subject faced by the present application. SUMMARY

[0007] The purpose of the present application is to provide a 2D-SLAM method and system for mobile robots in a bulk material yard.

[0008] In order to achieve the above-mentioned application purpose, the technical scheme adopted by the present application is as follows:

[0009] A 2D-SLAM method for mobile robots in a bulk material yard, comprising:

[0010] Step 1: Environmental Information Scan. Before starting work, the bulk material mobile robot scans the scrap yard environment with far-field lidar to determine the location of the bulk material pile and the drivable area.

[0011] Step 2: Global map construction. A global raster map of the bulk material storage yard is built using a map construction method based on the Cartographer framework.

[0012] Step 3: Obstacle avoidance map generation. When the bulk material mobile robot is working, it scans the environmental information around the material grabber with near-field LiDAR and generates a local obstacle avoidance map in real time.

[0013] Step two, the construction of the global map, includes the following steps:

[0014] a. Front-end scanning: The front-end scanning is responsible for adding the point bundles scanned in the space to the currently constructed subgraph;

[0015] b. Submap construction: A submap is part of the global raster map. A submap is constructed by creating several consecutive front-end scan results. The submap consists of probabilistic raster at a specified resolution. It uses a preset confidence probability threshold to distinguish each scanned point cloud data into three cases: obstructed, unknown, and unobstructed.

[0016] c. Scan matching: Scan matching is responsible for inserting the point cloud data scanned each time into the current sub-graph. By optimizing the position, angle and other information of the point cloud data, it inserts it into the optimal position under the current sub-graph.

[0017] In step a, the front-end scanning is responsible for adding the point bundles scanned in space to the currently constructed subgraph, and the pose of each point bundle can be represented by equation (1):

[0018] ξ=(ξ x ξ y ξ θ (1)

[0019] Where, ξ x Let ξ be the x-coordinate of the current point's endpoint. y Let ξ be the y-coordinate of the current point's endpoint. θ The angular coordinates of the current point's endpoint;

[0020] For a single lidar scan, a point cloud set H consisting of point beams can be obtained, as shown in equation (2):

[0021] H = {h} k} k＝1，......，K (2)

[0022] Among them, h kK represents a point beam in the point cloud set, and K represents the number of point clouds scanned once;

[0023] When a new scan is acquired, the conversion formula of the point cloud set H in the subgraph position is shown in formula (3):

[0024]

[0025] wherein T ξ represents the position matrix of the point cloud set H in the subgraph, and p represents the confidence probability of the current scan; for any scan, T ξ is obtained by solving, and the position information and confidence probability of all point beams under the current scan are obtained.

[0026] The subgraph in step b is part of the global grid map, and a subgraph is created by several consecutive front-end scan results. The subgraph is composed of probability grids at a specified resolution, which is divided into three cases of obstacles, unknown situations, and no obstacles by a preset confidence probability threshold, and its expression is shown in formula (4):

[0027] M: rZ*rZ→[p min , p max ] (4)

[0028] wherein M represents the constructed grid map, r represents the specified scan resolution, Z represents the point cloud data, and p min represents the nearest distance to the effective area of the mobile robot, and p max represents the farthest distance to the effective area of the mobile robot.

[0029] In step c, the scan matching is responsible for inserting each scanned point cloud data into the current subgraph, and the position, angle and other information of the point cloud data are optimized and inserted into the position under the current subgraph. The cost function of step c can be represented by formula (5):

[0030]

[0031] wherein M smooth is a linear evaluation function using bicubic interpolation.

[0032] The 2D-SLAM system of the mobile robot in the bulk material yard includes a data acquisition module and a map construction module. The data acquisition module includes a lower computer, a far-field laser radar, a near-field laser radar, an IMU and an odometer. The output ends of the far-field laser radar, the near-field laser radar, the IMU and the odometer are connected with the input end of the lower computer. The output end of the lower computer is connected with the input end of the map construction module.

[0033] The RS485 communication module is connected between the lower computer and the map construction module.

[0034] The map construction module is based on the ROS platform as a whole.

[0035] Compared with the prior art, the present application has the following beneficial effects:

[0036] 1. The present application is aimed at the SLAM construction problem of iron ore, coal, scrap iron and other bulk materials, and adopts a combination method of far-field laser radar and near-field laser radar for construction, wherein the far-field laser radar is used for rough construction, and the near-field laser radar is used for fine construction.

[0037] 2. The present application adopts a map construction method based on the Cartographer framework, improves the fineness of the construction, and solves the problem that the features scanned by the laser radar are prone to error matching due to the irregular placement of scrap steel piles.

[0038] 3. The present application adopts a combination of global grid map and local obstacle avoidance map, improves the obstacle avoidance sensing ability of the grabber under the premise of ensuring complete map information. BRIEF DESCRIPTION OF DRAWINGS

[0039] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application together with the embodiments thereof, and explain the present application, but do not constitute a limitation of the present application.

[0040] Figure 1 The present application is a 2D-SLAM method flowchart for a bulk material yard mobile robot.

[0041] Figure 2 The present application is a flowchart of Local SLAM under the Cartographer framework.

[0042] Figure 3 The present application is a flowchart of Global SLAM under the Cartographer framework.

[0043] Figure 4 The present application is a 2D-SLAM system architecture diagram for a mobile robot. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. Of course, the specific embodiments described herein are only used to explain the present application, and do not limit the present application.

[0045] Embodiment 1

[0046] Reference Figures 1-4 The technical scheme provided by the present application is:

[0047] A bulk material yard mobile robot 2D-SLAM method, comprising:

[0048] Step one, environment information scanning, before starting work, the bulk material mobile robot scans the scrap steel yard environment through the far-field laser radar, determines the bulk material pile position and the drivable area;

[0049] Step two, global map construction, a global grid map of the bulk material yard is established by using a map construction method based on the Cartographer framework;

[0050] Step three, obstacle avoidance map generation, when the bulk material mobile robot works, the environment information around the grabber is scanned by the near-field laser radar, and a local obstacle avoidance map is generated in real time.

[0051] The global map construction in step two includes the following steps:

[0052] a. Front-end scanning, the front-end scanning is responsible for adding the point beam scanned in space to the currently constructed subgraph;

[0053] b. Subgraph construction, the subgraph is a part of the global grid map, and the construction of a subgraph is created by several continuous front-end scanning results. The subgraph is composed of probability grids under a specified resolution, and each scanned point cloud data is divided into three cases of obstacles, unknown situations and no obstacles through a preset confidence probability threshold;

[0054] c. Scan matching, the scan matching is responsible for inserting each scanned point cloud data into the current subgraph, and the position, angle and other information of the point cloud data are optimized and inserted into the optimal position in the current subgraph.

[0055] The front-end scanning in step a is responsible for adding the point beam scanned in space to the currently constructed subgraph. The pose of each point beam can be expressed as formula (1):

[0056] ξ=(ξ x ,ξ y ,ξ θ ) (1)

[0057] Wherein, ξ x is the x coordinate of the current point beam end point, ξ y is the y coordinate of the current point beam end point, and ξ θ is the angle coordinate of the current point beam end point;

[0058] For one laser radar scanning, a point cloud set H composed of point beams can be obtained, which is expressed as formula (2):

[0059] H={h k} k＝1，......，K(2)

[0060] wherein, h k represents a point beam in the point cloud set, K represents the number of point clouds scanned once;

[0061] When a new scan is obtained, the conversion formula of the point cloud set H in the subgraph position is shown in formula (3):

[0062]

[0063] wherein, T ξ represents the position matrix of the point cloud set H in the subgraph, p represents the confidence probability of the current scan; for any scan, T ξ p item, the position information and confidence probability of all point beams under the current scan can be obtained.

[0064] The subgraph in step b is part of the global grid map, and a subgraph is created by several consecutive front-end scan results. The subgraph is composed of probability grids at a specified resolution, which distinguishes each scanned point cloud data into three cases of obstacles, unknown situations and no obstacles through a preset confidence probability threshold, and its expression is shown in formula (4):

[0065] M: rZ*rZ→[p min , p max ] (4)

[0066] wherein, M represents the constructed grid map, r represents the specified scan resolution, Z represents the point cloud data, represents the nearest distance to the effective area of the mobile robot, and represents the farthest distance to the effective area of the mobile robot.

[0067] In step c, the scan matching is responsible for inserting each scanned point cloud data into the current subgraph, and through the optimization processing of the position, angle and other information of the point cloud data, it is inserted into the position under the current subgraph. The cost function of this step c can be represented as formula (5):

[0068]

[0069] wherein, M smooth is a linear evaluation function using bicubic interpolation.

[0070] The 2D-SLAM system of the mobile robot in the bulk material yard includes a data acquisition module and a map construction module. The data acquisition module includes a lower computer, a far-field laser radar, a near-field laser radar, an IMU and an odometer. The output ends of the far-field laser radar, the near-field laser radar, the IMU and the odometer are connected with the input end of the lower computer. The output end of the lower computer is connected with the input end of the map construction module.

[0071] The lower computer and the map building module are connected by an RS485 communication module.

[0072] The map building module is based on a ROS platform as a whole.

[0073] Embodiment 2

[0074] On the basis of embodiment 1, the mobile robot 2D-SLAM method flow is as shown in Figure 1 which specifically includes the following three steps:

[0075] Step 1: Scanning the scrap yard environment by a far-field laser radar to determine the position of the bulk material pile and the drivable area.

[0076] Step 1 is implemented before the operation of the grabber, and its purpose is to pre-scan the scrap yard environment to provide scanning data for the subsequent establishment of the global grid map of the scrap yard. Among them, the far-field laser radar is used in this example to collect large-range point cloud data and determine the position of the scrap material pile and the drivable area, providing navigation information support for the movement of the grabber in the mapping process.

[0077] Step 2: Using a map building method based on the Cartographer framework to establish a global grid map of the scrap yard.

[0078] The data information of step 2 comes from the information collected by the far-field laser radar in step 1 and the pose information provided by the IMU and odometry sensors of the grabber. The map building method based on the Cartographer framework can be divided into two steps, namely Local SLAM and Global SLAM.

[0079] Local SLAM, namely local SLAM, is responsible for the establishment of a subgraph. Figure 2 The flowchart of Local SLAM is as follows: first, the input laser point cloud data is processed by two times of voxel filtering to filter the number of collected point clouds to the vicinity of the preset number, reducing the complexity of subsequent calculation; second, the point cloud data after voxel filtering is subjected to front-end scan matching, and the point cloud data collected by each scan is inserted into the optimal position of the subgraph; finally, the current laser point cloud data is inserted into the subgraph, and when the subgraph construction requirement is met, the data insertion is stopped, and the subgraph construction is completed.

[0080] Global SLAM, namely global SLAM, is responsible for rearranging and connecting each subgraph established by Local SLAM to form a coherent global graph. Figure 3is a flowchart of Global SLAM, firstly, Global SLAM acquires a subgraph of which scanning is completed and a node scanned currently; secondly, through loop closure detection, the position of the node scanned currently in the subgraph is determined, when the node scanned currently exists in multiple subgraphs, then the subgraphs are matched according to the position of the node in the subgraph through a graph optimization method, the graph optimization method is realized based on the ceres library of the Cartographer framework; when all the subgraphs are connected, a global map is output, and the global map construction is completed.

[0081] Step 3 scans the environment information around the grabber through the near-field laser radar to generate a local obstacle avoidance map in real time.

[0082] The implementation of step 3 is in the working process of the grabber, wherein the near-field laser radar is only used to acquire point cloud data of the environment around the grabber at a short distance, so as to establish a high-precision local obstacle avoidance map. The local obstacle avoidance map is generated and changed in real time, and the local obstacle avoidance map contains obstacles, scrap steel piles, other grabbers and other objects that can affect the operation and movement of the grabber.

[0083] The mobile robot 2D-SLAM system architecture is shown in Figure 4 The system comprises a data acquisition module and a map construction module.

[0084] The data acquisition module comprises a lower computer, a near-field laser radar, a far-field laser radar, an IMU and an odometer. The lower computer is directly connected with the above-mentioned various sensors, directly acquires the data of the sensors and performs preliminary processing, converts the data into data readable by the upper computer and transmits the processed data to the upper computer in real time. The data acquisition module provides system support for step 1 of the mobile robot 2D-SLAM method.

[0085] The map construction module is based on the ROS platform (Robot Operating System, ROS) as a whole, adopts the RS485 communication protocol to communicate with the lower computer and acquires the data required for map construction. The map construction module provides system support for steps 2 and 3 of the mobile robot 2D-SLAM method. The map construction module receives various sensor data processed by the data acquisition module, and completes the pre-construction of the scrap steel yard global map and the real-time construction of the local obstacle avoidance map.

[0086] The above only describes the preferred embodiments of the present application and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A 2D-SLAM method for mobile robots in bulk material storage yards, characterized in that, The method includes: Step 1: Environmental Information Scan. Before starting work, the bulk material mobile robot scans the scrap yard environment with far-field lidar to determine the location of the bulk material pile and the drivable area. Step 2: Global map construction. A global raster map of the bulk material storage yard is built using a map construction method based on the Cartographer framework. Step 3: Obstacle avoidance map generation. When the bulk material mobile robot is working, it scans the environmental information around the material grabber with near-field LiDAR and generates a local obstacle avoidance map in real time. Step two, the construction of the global map, includes the following steps: a. Front-end scanning: The front-end scanning is responsible for adding the point bundles scanned in the space to the currently constructed subgraph; b. Submap construction: A submap is part of the global raster map. A submap is constructed by creating several consecutive front-end scan results. The submap consists of probabilistic raster at a specified resolution. It uses a preset confidence probability threshold to distinguish each scanned point cloud data into three cases: obstructed, unknown, and unobstructed. c. Scan matching: Scan matching is responsible for inserting the point cloud data scanned each time into the current sub-graph. By optimizing the position, angle and other information of the point cloud data, it inserts it into the optimal position under the current sub-graph.

2. The 2D-SLAM method for a mobile robot in a bulk material storage yard according to claim 1, characterized in that, In step a, the front-end scanning is responsible for adding the point bundles scanned in space to the currently constructed subgraph, and the pose of each point bundle is represented as shown in equation (1):

3. The 2D-SLAM method for a mobile robot in a bulk material storage yard according to claim 1, characterized in that, In step b, the submap is part of the global raster map. A submap is constructed by creating several consecutive front-end scan results. The submap consists of probabilistic rasteres at a specified resolution. It distinguishes each scanned point cloud data into three cases: obstructed, unknown, and unobstructed, using a preset confidence probability threshold. Its expression is shown in equation (4):

4. The 2D-SLAM method for a mobile robot in a bulk material storage yard according to claim 1, characterized in that, In step c, the scan matching is responsible for inserting the point cloud data scanned each time into the current sub-image. By optimizing the position, angle and other information of the point cloud data, the data is inserted into the position under the current sub-image. The cost function of step c can be expressed as equation (5):

5. A 2D-SLAM system for a bulk material storage yard mobile robot using the method described in any one of claims 1-4, characterized in that, The 2D-SLAM system for mobile robots in bulk material yards includes a data acquisition module and a map building module. The data acquisition module includes a lower-level machine, a far-field lidar, a near-field lidar, an IMU, and an odometer. The output terminals of the far-field lidar, near-field lidar, IMU, and odometer are all connected to the input terminal of the lower-level machine, and the output terminal of the lower-level machine is connected to the input terminal of the map building module. The global map building module can implement the following steps of claim 1: front-end scanning, sub-map construction, and scan matching.

6. A 2D-SLAM system for a bulk material yard mobile robot according to claim 5, characterized in that, The lower-level machine and the map building module are connected via an RS485 communication module.

7. A 2D-SLAM system for mobile robots in bulk material storage yards according to claim 5, characterized in that, The map building module is based on the ROS platform.

Images