A multi-laser global positioning method and system

Abstract

This invention proposes a multi-laser global positioning method and system. The method includes: after calibrating the robot's laser radar external parameters, using the calibrated laser radar to fuse and scan the indoor structured environment to obtain three-dimensional point cloud data; using historical three-dimensional point cloud data to construct a map of the indoor structured environment; extracting edge feature points from the historical three-dimensional point cloud data to obtain the corresponding position and posture of the edge feature points in the indoor structured environment map; matching the acquired real-time laser point cloud data with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data to obtain the current robot's posture information. Based on this method, a multi-laser global positioning system is also proposed. The present invention realizes re-positioning of the robot in the structured environment of a ship cabin after power is cycled or the position and posture suddenly change, solving the problems of re-positioning in the absence of prior knowledge and loop detection and global positioning of the robot due to environmental changes, effectively improving the stability and robustness of positioning.

Description

Technical Field

[0001] The present invention belongs to the technical field of navigation and positioning, and in particular relates to a multi-laser global positioning method and system. Background Art

[0002] Intelligent ports and smart operations have become a key trend in port construction and development. Fully hydraulic crawler-type tank cleaning machines that automatically perform tank cleaning operations can save time, reduce labor costs, and significantly enhance the ability to optimize the allocation of port-related resources. Accurate and stable positioning is a prerequisite for autonomous navigation. In unmanned driving, global positioning generally refers to positioning on a pre-built map after powering on or after the robot is actuated. At this point, the current frame's pose may appear anywhere within the existing map, or even outside of it.

[0003] Generally speaking, it means that without prior information, the robot relies solely on its own sensors to estimate its position and posture in a known global map. There are three main traditional methods: one is to use RTK (carrier phase difference technology) as global constraint information to add to the positioning; the second is to use key frames to find key frames that match the current frame from the known map, and the position and posture of the key frames are known; the third is to directly use all the point clouds of the current frame and the map for brute force search and matching. The difficulties and challenges of global positioning come from two aspects: (1) lack of prior knowledge; that is, the position of the robot at the previous moment is unknown; (2) when re-localizing and closed-loop detection, that is, when the robot returns to a location that has been visited before, various changes will occur, including changes in lighting and perspective, as well as changes caused by some dynamic objects.

[0004] Positioning is the first step toward autonomous driving. Currently, mainstream SLAM positioning methods include gmapping, NDT, and cartographer. During navigation, external factors (such as human removal or collisions) can cause the robot's position to change suddenly, rendering existing positioning algorithms that rely on continuous position changes ineffective. In indoor environments, integrated navigation loses GNSS signals, making it impossible to establish global constraints. While the keyframe method is faster, keyframes in similar environments are not only redundant but also prone to mismatching. Direct laser matching is also slower and unsuitable for global positioning in larger scenarios. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes a multi-laser global positioning method and system, which effectively solves the repositioning problem in the absence of prior knowledge and the loop detection and global positioning problem when the robot returns to a previously visited location due to environmental changes, effectively improving the stability and robustness of positioning.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A multi-laser global positioning method comprises the following steps:

[0008] After calibrating the robot's lidar external parameters, the calibrated lidar is used to fusion scan the indoor structured environment to obtain 3D point cloud data;

[0009] Use historical 3D point cloud data for SLAM mapping to build indoor structured environment maps;

[0010] Edge feature points are extracted from historical three-dimensional point cloud data, and the corresponding positions of the edge feature points are obtained in the constructed indoor structured environment map; the acquired real-time laser point cloud data is matched with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data to obtain the current robot's posture information.

[0011] Furthermore, the method for calibrating the external parameters of the robot's lidar includes: using the main lidar coordinate system located directly above the robot as a reference, transforming the right hemisphere radar coordinate system and the left hemisphere radar coordinate system into the main lidar coordinate system; the right hemisphere radar coordinate system is located on the right side of the robot; and the left hemisphere radar coordinate system is located on the left side of the robot.

[0012] Furthermore, the indoor structured environment includes a cabin environment.

[0013] Furthermore, the process of constructing the indoor structured environment map includes: the robot is equipped with an inertial measurement unit and a lidar to scan the cabin environment, while saving the sensor data packets, adapting the cartographer algorithm according to the parameters of the lidar and the inertial measurement unit to build the map, and obtaining the binary form of the map.

[0014] Furthermore, the process of extracting edge feature points from historical three-dimensional point cloud data includes: extracting feature points based on curvature, dividing a laser line into multiple blocks evenly in order to have uniform constraints on a circle, arranging the points in the blocks according to the size of the curvature, and taking the point with the largest curvature as the edge feature point, and the point with the smallest curvature as the plane point.

[0015] Furthermore, the curvature is calculated as follows:

[0016]

[0017] Among them, i is the historical three-dimensional point cloud data P k A point in k; S is the continuous point set of i returned by the laser scanner in the same scan, and j is the point set around i; the laser radar coordinate system L is a three-dimensional coordinate system with its origin at the geometric center of the laser radar; the x-axis points to the left, the y-axis points upward, and the z-axis points forward; the laser point i∈P k , in L k The coordinate system is expressed as

[0018] Furthermore, the process of matching the acquired real-time laser point cloud data with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data includes: using feature matching to determine whether the distance function between the real-time laser point cloud data and the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data is less than a threshold; if it is less than the threshold, the position of the real-time laser point cloud data is the same as that of the edge feature point.

[0019] Furthermore, the process of determining the current posture of the robot includes:

[0020] In the constructed indoor structured environment map, the edge feature points are represented as Represents the first feature point in the indoor structured environment map; Represents the second feature point in the indoor structured environment map; Represents the Nth feature point in the indoor structured environment map;

[0021] The position and posture corresponding to the edge feature point in the constructed indoor structured environment map is expressed as Represents the pose of the first feature point in the indoor structured environment map; Represents the pose of the second feature point in the indoor structured environment map; Represents the pose of the Nth feature point in the indoor structured environment map;

[0022] The matching symbol corresponding to the acquired real-time laser point cloud data is represented as I q ; Through feature matching, determine the frame number k with the minimum distance function in I. If the distance function is less than the threshold, the current posture T q The expression is:

[0023]

[0024] Among them, R is the rotation matrix; t is the translation matrix, which is specifically expressed as:

[0025]

[0026] t=[0 0 0] T; θ is the heading angle offset obtained from edge feature point matching.

[0027] The present invention also proposes a multi-laser global positioning system, comprising a calibration module, a mapping module and a positioning module;

[0028] The calibration module is used to calibrate the external parameters of the robot laser radar and use the calibrated laser radar to fuse and scan the indoor structured environment to obtain three-dimensional point cloud data;

[0029] The mapping module is used to perform SLAM mapping using historical three-dimensional point cloud data to construct an indoor structured environment map;

[0030] The positioning module is used to extract edge feature points from historical three-dimensional point cloud data and obtain the position and posture corresponding to the edge feature points in the constructed indoor structured environment map; the acquired real-time laser point cloud data is matched with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data to obtain the current robot's posture information.

[0031] Furthermore, the process implemented by the mapping module is as follows: when the indoor structured environment is a cabin environment, the robot is equipped with an inertial measurement unit and a lidar to scan the cabin environment, while saving the sensor data packets, and adapting the cartographer algorithm according to the parameters of the lidar and the inertial measurement unit to obtain a binary form of the map.

[0032] The effects provided in the summary of the invention are only the effects of the embodiments, not all the effects of the invention. One of the above technical solutions has the following advantages or beneficial effects:

[0033] The present invention proposes a multi-laser global positioning method and system, which includes the following steps: after calibrating the robot's laser radar extrinsic parameters, using the calibrated laser radar to fuse and scan an indoor structured environment to obtain three-dimensional point cloud data; using historical three-dimensional point cloud data for SLAM mapping to construct a map of the indoor structured environment; extracting edge feature points from the historical three-dimensional point cloud data, and obtaining the corresponding poses of the edge feature points in the constructed indoor structured environment map; matching the acquired real-time laser point cloud data with descriptors encoded into edge feature points in the historical three-dimensional point cloud data to obtain the current robot's posture information. Based on a multi-laser global positioning method, a multi-laser global positioning system is also proposed. The present invention first scans the environment through multi-laser radar data fusion to obtain a three-dimensional point cloud set, uses this historical three-dimensional point cloud for SLAM mapping, and extracts edge feature points from this point cloud; during positioning, edge feature descriptors extracted in real time from the current frame point cloud are matched with the aforementioned historical feature descriptors to obtain the current robot's posture.

[0034] The present invention can realize the repositioning of the robot after it is powered on again or its posture changes suddenly in the structured environment of the cabin, effectively solving the repositioning problem in the absence of prior knowledge and the loop detection and global positioning problems caused by environmental changes when the robot returns to a previously visited location, and effectively improving the stability and robustness of positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] like Figure 1 This is a flow chart of a multi-laser global positioning method proposed in Example 1 of the present invention;

[0036] like Figure 2 This is a schematic diagram of the installation of laser radars around the robot in Example 1 of the present invention;

[0037] like Figure 3 This is a schematic diagram of the laser radar extrinsic calibration proposed in Example 1 of the present invention;

[0038] like Figure 4 Schematic diagram of edge feature point conditions proposed in Example 1 of the present invention;

[0039] like Figure 5 This is a schematic diagram of the edge feature points extracted according to Example 1 of the present invention;

[0040] like Figure 6 This is a schematic diagram of a multi-laser global positioning system proposed in Example 2 of the present invention. DETAILED DESCRIPTION

[0041] In order to clearly illustrate the technical features of this solution, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings. The disclosure below provides many different embodiments or examples for realizing different structures of the present invention. In order to simplify the disclosure of the present invention, the components and settings of specific examples are described below. In addition, the present invention may repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplicity and clarity and does not itself indicate the relationship between the various embodiments and / or settings discussed. It should be noted that the components illustrated in the accompanying drawings are not necessarily drawn to scale. The present invention omits descriptions of well-known components and processing technologies and processes to avoid unnecessary limitations on the present invention.

[0042] Example 1

[0043] Embodiment 1 of the present invention proposes a multi-laser global positioning method to solve the re-positioning problem in the absence of prior knowledge, as well as the loop detection and global positioning problem of the robot returning to a previously visited location due to environmental changes. Figure 1 This is a flow chart of a multi-laser global positioning method proposed in Example 1 of the present invention.

[0044] Calibrate the robot's lidar external parameters. The method for calibrating the robot's lidar external parameters includes: using the main lidar coordinate system located directly above the robot as a reference, transforming the right hemisphere radar coordinate system and the left hemisphere radar coordinate system into the main lidar coordinate system; Figure 2 Schematic diagram of the installation of laser radars around the robot in Example 1 of the present invention; the right hemisphere radar coordinate system is located on the right side of the robot; the left hemisphere radar coordinate system is located on the left side of the robot.

[0045] like Figure 3 This is a schematic diagram of the laser radar external parameter calibration proposed in Example 1 of the present invention; if the robot uses multiple laser radars, the external parameters need to be calibrated to facilitate subsequent point cloud fusion, etc. The laser radar external parameter calibration is essentially to obtain the displacement (x, y, z) and rotation (roll, pitch, yaw) of the two lasers. Generally, the main laser is used as a reference (parent coordinate system), and the other two side laser radars (child coordinate systems) are transformed into the coordinate system of the main laser. This transformation relationship is described in three-dimensional space by the secondary transformation matrix (R, T) (Homogeneous transformation matrix). After the calibration is completed, the static TF of the main laser radar and the side laser radar are released.

[0046] Historical 3D point cloud data is used for SLAM mapping to construct an indoor structured environment map. First, the three lidars are synchronized in time, and preprocessed such as anomaly judgment and dedistortion are performed. After the processing is completed, the point clouds are merged using the TF coordinate transformation obtained in the previous step, and the merged point cloud data is published.

[0047] This paper proposes a multi-laser global positioning method for structured indoor environments, suitable for ship cabins. The process of building a cabin environment map involves: a robot equipped with an inertial measurement unit (IMU) and a lidar (LiDAR) scans the cabin environment, saving the sensor data packets, and adapting the Cartographer algorithm based on the LiDAR and IMU parameters to create a map, generating a binary .pbstream representation of the map.

[0048] Extract edge feature points from historical 3D point cloud data. The number of feature points is very small compared to all point clouds, so the speed and performance of subsequent matching can be greatly improved. Traverse the curvature parameters from large to small, so that the largest jump point can be traversed first, that is, the edge feature point - corner point (line feature). Figure 4 This is a schematic diagram of the edge feature point conditions proposed in Example 1 of the present invention; the conditions for the edge feature point must not be an edge point that jumps from a distant point to a nearby point, must not be a point on the ground, must have a smoothness greater than a certain value, and the five adjacent points of the feature point must not be feature points; that is, if one point is a corner point, the next five adjacent points must not be corner points.

[0049] Only the first 20 points in each sample (i.e. the 20 points with the largest curvature) are extracted. Figure 5 This is a schematic diagram of edge feature points extracted from the entire environment proposed in Example 1 of the present invention.

[0050] The curvature is calculated as:

[0051]

[0052] Among them, i is the historical three-dimensional point cloud data P k A point in k ; S is the continuous point set of i returned by the laser scanner in the same scan, and j is the point set around i; the laser radar coordinate system L is a three-dimensional coordinate system with its origin at the geometric center of the laser radar; the x-axis points to the left, the y-axis points upward, and the z-axis points forward; the laser point i∈P k , in L k The coordinate system is expressed as

[0053] Feature points are extracted based on curvature. In order to ensure uniform constraints over 360 degrees, a laser line is divided into 6 blocks. The points in the blocks are arranged according to the size of the curvature. The two points with the largest curvature are used as edge feature points (cornersharp points), and the four points with the smallest curvature are used as plane points (surf flat points).

[0054] After Cartographer completes the mapping of a specific environment, the position and pose of each frame of laser point cloud collected by the lidar during mapping is determined. By matching it with the edge points (corner points), positioning in the environment can be completed.

[0055] The process of matching the acquired real-time laser point cloud data with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data includes: using feature matching to determine whether the distance function between the real-time laser point cloud data and the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data is less than a threshold; if it is less than the threshold, the position of the real-time laser point cloud data is the same as that of the edge feature point.

[0056] The work of recording and saving the edge feature descriptors and poses corresponding to the map construction has been completed before the positioning begins. This work is collectively referred to as library construction. During actual positioning, you only need to call the results in the library.

[0057] The process of determining the current robot pose includes:

[0058] In the constructed indoor structured environment map, the edge feature points are represented as Represents the first feature point in the indoor structured environment map; Represents the second feature point in the indoor structured environment map; Represents the Nth feature point in the indoor structured environment map;

[0059] The position and posture corresponding to the edge feature point in the constructed indoor structured environment map is expressed as Represents the pose of the first feature point in the indoor structured environment map; Represents the pose of the second feature point in the indoor structured environment map; Represents the pose of the Nth feature point in the indoor structured environment map;

[0060] The matching symbol corresponding to the acquired real-time laser point cloud data is represented as I q ; Through feature matching, determine the frame number k with the minimum distance function in I. If the distance function is less than the threshold, the current posture T q The expression is:

[0061]

[0062] Among them, R is the rotation matrix; t is the translation matrix, which is specifically expressed as:

[0063]

[0064] t=[0 0 0] T ; θ is the heading angle offset obtained from edge feature point matching.

[0065] The initial pose is obtained through the above feature matching, and then a more accurate initial pose is obtained through the ICP or NDT matching method.

[0066] A multi-laser global positioning method proposed in Example 1 of the present invention first obtains a three-dimensional point cloud set by fusing multi-lidar data to scan the environment, uses the historical three-dimensional point cloud for SLAM mapping, and extracts edge feature points through the point cloud; during positioning, the edge feature descriptor obtained by real-time extraction of the current frame point cloud is feature matched with the above-mentioned historical feature descriptors to obtain the current robot's posture.

[0067] A multi-laser global positioning method proposed in Example 1 of the present invention can achieve repositioning of the robot after it is powered on again or its posture suddenly changes in the structured environment of the cabin, effectively solving the repositioning problem in the absence of prior knowledge and the loop detection and global positioning problem of the robot returning to a previously visited location due to environmental changes, effectively improving the stability and robustness of positioning.

[0068] Example 2

[0069] Based on a multi-laser global positioning method proposed in Example 1 of the present invention. A multi-laser global positioning system proposed in Example 2 of the present invention, such as Figure 6 This is a schematic diagram of a multi-laser global positioning system proposed in Example 2 of the present invention, which includes a calibration module, a mapping module, and a positioning module;

[0070] The calibration module is used to calibrate the external parameters of the robot's lidar and then use the calibrated lidar to fuse and scan the indoor structured environment to obtain three-dimensional point cloud data;

[0071] The mapping module is used to perform SLAM mapping using historical 3D point cloud data to construct indoor structured environment maps;

[0072] The positioning module is used to extract edge feature points from historical three-dimensional point cloud data and obtain the corresponding position and posture of the edge feature points in the constructed indoor structured environment map; the acquired real-time laser point cloud data is matched with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data to obtain the current robot's posture information.

[0073] The calibration module implementation process includes: using the main lidar coordinate system located directly above the robot as a reference, transforming the right hemisphere radar coordinate system and the left hemisphere radar coordinate system into the main lidar coordinate system; the right hemisphere radar coordinate system is located on the right side of the robot; and the left hemisphere radar coordinate system is located on the left side of the robot.

[0074] If a robot uses multiple lidars, external parameter calibration is required to facilitate subsequent point cloud fusion and other tasks. LiDAR external parameter calibration essentially involves obtaining the displacement (x, y, z) and rotation (roll, pitch, yaw) of the two lasers. Typically, the primary laser is used as a reference (the parent coordinate system) and the other two side lidars (child coordinate systems) are transformed to the primary laser's coordinate system. This transformation is described in three-dimensional space using the secondary transformation matrix (R, T). After calibration, the static TFs of the primary and side lidars are published.

[0075] The mapping module implementation process includes: using historical 3D point cloud data for SLAM mapping to construct a structured indoor environment map. First, the three lidars are time-synchronized, and pre-processed such as anomaly judgment and distortion removal are performed. After processing is completed, the point clouds are merged using the TF coordinate transformation obtained in the previous step, and the merged point cloud data is published.

[0076] When the indoor structured environment is a cabin environment, the robot is equipped with an inertial measurement unit and lidar to scan the cabin environment, while saving the sensor data packets. The cartographer algorithm is adapted and mapped according to the parameters of the lidar and the inertial measurement unit to obtain a binary form of the map.

[0077] The localization module implementation process involves extracting edge feature points from historical 3D point cloud data. The extremely small number of feature points relative to the entire point cloud significantly improves the speed and performance of subsequent matching. The curvature parameters are then traversed from largest to smallest, initially traversing to the largest jump points, which are edge feature points—corner points (line features). Edge feature points must not be edge points that jump from distant locations to nearby locations, must not be points on the ground, must have a smoothness greater than a certain value, and must not be adjacent to any of the five points that are also feature points. This means that if one point is a corner point, the next five adjacent points must also not be corner points.

[0078] Only the first 20 points in each sample (i.e. the 20 points with the largest curvature) are extracted. Figure 5 This is a schematic diagram of edge feature points extracted from the entire environment proposed in Example 1 of the present invention.

[0079] The curvature is calculated as:

[0080]

[0081] Among them, i is the historical three-dimensional point cloud data P k A point in k ; S is the continuous point set of i returned by the laser scanner in the same scan, and j is the point set around i; the laser radar coordinate system L is a three-dimensional coordinate system with its origin at the geometric center of the laser radar; the x-axis points to the left, the y-axis points upward, and the z-axis points forward; the laser point i∈P k , in L k The coordinate system is expressed as

[0082] Feature points are extracted based on curvature. In order to ensure uniform constraints over 360 degrees, a laser line is divided into 6 blocks. The points in the blocks are arranged according to the size of the curvature. The two points with the largest curvature are used as edge feature points (cornersharp points), and the four points with the smallest curvature are used as plane points (surf flat points).

[0083] After Cartographer completes the mapping of a specific environment, the position and pose of each frame of laser point cloud collected by the lidar during mapping is determined. By matching it with the edge points (corner points), positioning in the environment can be completed.

[0084] The process of matching the acquired real-time laser point cloud data with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data includes: using feature matching to determine whether the distance function between the real-time laser point cloud data and the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data is less than a threshold; if it is less than the threshold, the position of the real-time laser point cloud data is the same as that of the edge feature point.

[0085] The work of recording and saving the edge feature descriptors and poses corresponding to the map construction has been completed before the positioning begins. This work is collectively referred to as library construction. During actual positioning, you only need to call the results in the library.

[0086] The process of determining the current robot pose includes:

[0087] In the constructed indoor structured environment map, the edge feature points are represented as Represents the first feature point in the indoor structured environment map; Represents the second feature point in the indoor structured environment map; Represents the Nth feature point in the indoor structured environment map;

[0088] The position and posture corresponding to the edge feature point in the constructed indoor structured environment map is expressed as Represents the pose of the first feature point in the indoor structured environment map; Represents the pose of the second feature point in the indoor structured environment map; Represents the pose of the Nth feature point in the indoor structured environment map;

[0089] The matching symbol corresponding to the acquired real-time laser point cloud data is represented as I q ; Through feature matching, determine the frame number k with the minimum distance function in I. If the distance function is less than the threshold, the current posture T q The expression is:

[0090]

[0091] Among them, R is the rotation matrix; t is the translation matrix, which is specifically expressed as:

[0092]

[0093] t=[0 0 0] T ; θ is the heading angle offset obtained from edge feature point matching.

[0094] The initial pose is obtained through the above feature matching, and then a more accurate initial pose is obtained through the ICP or NDT matching method.

[0095] A multi-laser global positioning system proposed in Example 2 of the present invention first obtains a three-dimensional point cloud set by fusing multi-lidar data to scan the environment, uses the historical three-dimensional point cloud for SLAM mapping, and extracts edge feature points through the point cloud; during positioning, the edge feature descriptor obtained by real-time extraction of the current frame point cloud is feature matched with the above-mentioned historical feature descriptors to obtain the current robot's posture.

[0096] A multi-laser global positioning system proposed in Example 2 of the present invention can achieve repositioning of the robot after it is powered on again or its posture suddenly changes in the structured environment of the cabin, effectively solving the repositioning problem in the absence of prior knowledge and the loop detection and global positioning problem of the robot returning to a previously visited location due to environmental changes, effectively improving the stability and robustness of positioning.

[0097] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device comprising a series of elements are inherent to the elements. In the absence of further restrictions, the elements limited by the statement "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device comprising the elements. In addition, the above-mentioned technical solutions provided in the embodiments of the present application are not described in detail in accordance with the corresponding technical solutions in the prior art to achieve the same principle, so as to avoid excessive elaboration.

[0098] Although the above description is of specific embodiments of the present invention in conjunction with the accompanying drawings, it does not limit the scope of protection of the present invention. For those skilled in the art, other different forms of modifications or variations can be made based on the above description. It is not necessary and impossible to list all embodiments here. Based on the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without expending creative effort are still within the scope of protection of the present invention.

Claims

1. A multi-laser global positioning method, characterized in that: The following steps are involved: After calibrating the robot's lidar external parameters, the calibrated lidar is used to fusion scan the indoor structured environment to obtain 3D point cloud data; Use historical 3D point cloud data for SLAM mapping to build indoor structured environment maps; Extracting edge feature points from historical three-dimensional point cloud data and obtaining the pose corresponding to the edge feature points in the constructed indoor structured environment map; matching the acquired real-time laser point cloud data with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data to obtain the current robot posture information; The process of extracting edge feature points from historical three-dimensional point cloud data includes: extracting feature points based on curvature, dividing a laser line into multiple blocks to ensure uniform constraints on a circle, arranging the points in the blocks according to the size of the curvature, and taking the point with the largest curvature as the edge feature point and the point with the smallest curvature as the plane point; The curvature is calculated as follows: ； in, Historical 3D point cloud data A point in ; The laser scanner returns the The continuous point set of for The surrounding point set; the laser radar coordinate system L is a three-dimensional coordinate system with its origin at the geometric center of the laser radar; the x-axis points to the left, the y-axis points upward, and the z-axis points forward; the laser point ,exist The coordinate system is expressed as ; The current robot posture information is: In the constructed indoor structured environment map, the edge feature points are represented as ; Represents the first feature point in the indoor structured environment map; Represents the second feature point in the indoor structured environment map; Represents the Nth feature point in the indoor structured environment map; The position and posture corresponding to the edge feature point in the constructed indoor structured environment map is expressed as ; Represents the pose of the first feature point in the indoor structured environment map; Represents the pose of the second feature point in the indoor structured environment map; Represents the pose of the Nth feature point in the indoor structured environment map; The matching symbol corresponding to the acquired real-time laser point cloud data is expressed as ; Through feature matching in The frame number of the smallest distance function is determined as k. If the distance function is less than the threshold, the current pose The expression is: ； Among them, R is the rotation matrix; t is the translation matrix, which is specifically expressed as: ; It is the heading angle offset obtained from edge feature point matching.

2. A multi-laser global positioning method according to claim 1, characterized in that: The method for calibrating the external parameters of the robot's lidar includes: using the main lidar coordinate system located directly above the robot as a reference, transforming the right hemisphere radar coordinate system and the left hemisphere radar coordinate system into the main lidar coordinate system; the right hemisphere radar coordinate system is located on the right side of the robot; and the left hemisphere radar coordinate system is located on the left side of the robot.

3. A multi-laser global positioning method according to claim 1, characterized in that: The indoor structured environment includes a ship cabin environment.

4. A multi-laser global positioning method according to claim 3, characterized in that: The process of constructing the indoor structured environment map includes: the robot is equipped with an inertial measurement unit and a lidar to scan the cabin environment, while saving the sensor data packets, and adapting the cartographer algorithm according to the parameters of the lidar and the inertial measurement unit to obtain a binary form of the map.

5. The multi-laser global positioning method according to claim 1, characterized in that: The process of matching the acquired real-time laser point cloud data with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data includes: using a feature matching method to determine whether a distance function between the real-time laser point cloud data and the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data is less than a threshold; if so, the real-time laser point cloud data and the edge feature points have the same position.

6. A multi-laser global positioning system for executing a multi-laser global positioning method according to any one of claims 1 to 5, characterized in that: Includes calibration module, mapping module and positioning module; The calibration module is used to calibrate the external parameters of the robot laser radar and use the calibrated laser radar to fuse and scan the indoor structured environment to obtain three-dimensional point cloud data; The mapping module is used to perform SLAM mapping using historical three-dimensional point cloud data to construct an indoor structured environment map; The positioning module is used to extract edge feature points from historical three-dimensional point cloud data and obtain the position and posture corresponding to the edge feature points in the constructed indoor structured environment map; the acquired real-time laser point cloud data is matched with the descriptors encoded into the edge feature points in the historical three-dimensional point cloud data to obtain the current robot's posture information.

7. A multi-laser global positioning system according to claim 6, characterized in that: The mapping module implements the following process: when the indoor structured environment is a cabin environment, the robot is equipped with an inertial measurement unit and a lidar to scan the cabin environment, while saving the sensor data packets, and adapting the cartographer algorithm according to the parameters of the lidar and the inertial measurement unit to obtain a binary form of the map.

Images