Global repositioning method, system and device for mobile robot and medium

Abstract

The invention discloses a global repositioning method, system and device for a mobile robot and a medium. The method comprises the steps that the robot obtains a current scanning frame through laser radar scanning; for the point cloud data of the current scanning frame, generating a BEV picture in a voxelization mode; then extracting local features of each BEV picture, and encoding the local features to generate global features; searching a global feature matched with the global feature of the current scanning frame in a feature library, and obtaining a local feature corresponding to the matched global feature and robot pose information; performing violent matching on the local features of the current scanning frame and the obtained local features to obtain corresponding points; performing pose estimation on the corresponding points to obtain a relative transformation matrix between the two local features; and then the relative transformation matrix acts on the obtained robot pose information to obtain the current global position information of the robot, so that the global positioning capability of the robot under the condition of losing positioning or having no initialized pose is improved.

Description

Technical Field

[0001] The invention relates to a fast and robust global repositioning method based on search, and in particular to a global repositioning method, system, device and medium for a mobile robot, belonging to the field of automation technology. Background Art

[0002] With the rapid development of mobile robot technology, its application in industrial automation, service industry, warehousing and logistics is becoming more and more extensive. Mobile robots usually need to autonomously navigate and locate in their working environment to complete specific tasks or operations. To achieve this goal, mobile robots are usually equipped with a variety of sensors, such as laser radar (LiDAR), inertial measurement unit (IMU) and wheel speed meter, etc., and obtain their position information in the global map in real time through simultaneous localization and mapping (SLAM) technology.

[0003] However, in practical applications, the mobile robot may lose positioning information or fail to initialize its position due to various reasons (such as sensor failure, environmental changes, signal interference, etc.), resulting in the robot being unable to accurately determine its position in the global map. In this case, the robot needs to perform global relocalization to regain its accurate position information so as to continue to perform the task.

[0004] The existing global positioning solutions mainly include the following: 1) ‌Location recognition only method‌: This method directly uses point clouds to generate global descriptors, and then retrieves the most similar frame in the database based on the descriptor distance to obtain the location estimate. However, this method has problems such as inaccurate feature expression, limited generalization ability, and being affected by scale distortion, resulting in insufficient location recognition accuracy in complex environments or new scenes.

[0005] 2) Global pose estimation method: This method uses local features of point clouds (such as geometric relationships, traditional image descriptors or segmentation information, etc.) to generate local descriptors, and directly determines the global pose by matching with the global map. However, this method is easily affected by environmental noise and point cloud density, and has problems such as unstable key point detection, low efficiency and poor adaptability.

[0006] 3) Local pose estimation after position recognition: This method first uses global descriptor retrieval to obtain a rough position, and then estimates the relative pose based on local feature matching to obtain the global pose. However, this method has problems such as insufficient generalization ability, feature extraction accuracy affected by point cloud sparsity, poor adaptability to data distribution changes, and unstable key point detection.

[0007] In summary, the existing global positioning methods have many problems in terms of point cloud sparsity, robustness to rotation changes, generalization ability, dependence on precise pose supervision, efficiency and resource requirements, etc., which make it difficult to meet the global positioning requirements of mobile robots in complex environments with high precision, robustness, efficiency and ease of deployment.

[0008] Therefore, a new global relocalization method for mobile robots is urgently needed to solve the above problems. Summary of the invention

[0009] In response to the above-mentioned existing technical problems, the present invention provides a mobile robot global repositioning method, system, device and medium to improve the global positioning capability of the mobile robot when it loses its positioning or has no initialized posture, thereby improving its reliability and efficiency in practical applications.

[0010] To achieve the above technical objectives, firstly, the present invention provides an automatic deployment system for a robot control system, including a mapping stage and a positioning stage.

[0011] The mapping stage includes the following steps: The robot uses the starting position of the mapping operation as the map origin, estimates the movement distance of the robot relative to the starting position through the odometer, and updates the robot's posture information in real time; every time the robot moves a fixed distance, the scan frame obtained by the current lidar is used as the key frame, and the current robot posture information obtained from the odometer is used as the robot posture of the key frame; For the point cloud data of each key frame, a BEV image is generated by voxelization; then, each BEV image is rotated to enhance the data and generate multiple rotated BEV images; then, the local features of each BEV image are extracted and global features are generated from the local features; The robot posture, BEV image, local features and global features corresponding to each key frame are bound and stored in the feature library; after the map is built, the point cloud data of the global map and the feature library composed of all key frames are obtained.

[0012] The positioning stage includes the following steps: The robot obtains the current scan frame through laser radar scanning; generates a BEV image by voxelization of the point cloud data of the current scan frame; then extracts the local features of each BEV image, and generates global features from the local features; Retrieve the global features that match the global features of the current scan frame in the feature library, and obtain the local features and robot posture information corresponding to the matched global features; The local features of the current scanning frame are matched with the acquired local features by brute force to obtain corresponding points; the corresponding points are estimated to obtain the relative transformation matrix between the local features of the current scanning frame and the acquired local features; the relative transformation matrix is ​​then applied to the acquired robot posture information to obtain the robot's current global position information.

[0013] In the method of the present invention, the fixed distance is 0.5 meters.

[0014] The method of the present invention further provides that the rotation data is enhanced by rotating 45 degrees each time to generate 8 rotated BEV images.

[0015] The method of the present invention further comprises extracting the local features through a residual network Resnet.

[0016] The method of the present invention further comprises generating the global features through NetVLAD network coding.

[0017] In the method of the present invention, the brute force matching adopts a feature point matching algorithm.

[0018] The method of the present invention further adopts a RANSAC algorithm for posture estimation.

[0019] Secondly, the present invention also provides a mobile robot global repositioning system, which utilizes the mobile robot global repositioning method and includes a mapping module and a positioning module.

[0020] The mapping module includes a key frame generation unit, a key frame processing unit, and a feature library generation unit; The key frame generation unit is used to: the robot takes the starting position of the map construction operation as the map origin, estimates the movement distance of the robot relative to the starting position through the odometer, and updates the robot's posture information in real time; each time the robot moves a fixed distance, the scanning frame obtained by the current laser radar is used as the key frame, and the current robot posture information is obtained from the odometer as the robot posture of the key frame; The key frame processing unit is used to: generate a BEV image by voxelization for the point cloud data of each key frame; then perform rotation data enhancement on each BEV image to generate multiple rotated BEV images; then extract local features of each BEV image, and generate global features from the local features; The feature library generation unit is used to: bind the robot posture, BEV image, local features and global features corresponding to each key frame, and store them in the feature library; after the map is built, the point cloud data of the global map and the feature library composed of all key frames are obtained.

[0021] The positioning module includes a global feature generation unit, a feature library retrieval unit, and a global position acquisition unit; The global feature generation unit is used to: the robot obtains the current scanning frame through laser radar scanning; generates a BEV picture by voxelization for the point cloud data of the current scanning frame; then extracts the local features of each BEV picture, and generates global features from the local features; The feature library retrieval unit is used to: retrieve the global features matching the global features of the current scanning frame in the feature library, and obtain the local features and robot posture information corresponding to the matching global features; The global position acquisition unit is used to: obtain corresponding points by brute force matching the local features of the current scanning frame with the acquired local features; and estimate the posture of the corresponding points to obtain the relative transformation matrix between the local features of the current scanning frame and the acquired local features; and then apply the relative transformation matrix to the acquired robot posture information to obtain the current global position information of the robot.

[0022] Thirdly, the present invention also provides a mobile robot global repositioning device, comprising at least one processor; and a memory connected to at least one of the processors; Wherein, the memory stores instructions executable by the processor; The instructions are used to be executed by the processor to implement the global repositioning method of a mobile robot.

[0023] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the mobile robot global repositioning method is implemented.

[0024] It can be seen from the above technical scheme that the present invention proposes a global relocalization method for a mobile robot based on generating a bird's-eye view (BEV) based on point cloud data. Through technical innovation, it solves the problems of point cloud sparsity, insufficient robustness to rotation changes, limited generalization ability and low efficiency existing in the existing global positioning methods, and has the following significant beneficial effects.

[0025] 1) Generate BEV images from point cloud data to improve feature extraction stability: The present invention converts point cloud data into BEV images by voxelization, which significantly reduces the impact of point cloud sparsity on feature extraction compared to the prior art method of directly extracting features based on radar points. BEV images can better preserve the global structural information of the scene while reducing the interference caused by changes in point cloud density, thereby improving the stability and accuracy of feature extraction.

[0026] 2) ‌Rotation data augmentation to enhance rotation change robustness and generalization ability‌: When generating global features in the mapping stage, the present invention performs rotational data enhancement on the input BEV images, rotating them 45 degrees each time, so that each BEV image is expanded to 8. This method not only increases the diversity of training data, but also significantly improves the robustness and generalization ability of the network to scene rotation changes. In this way, the network can better adapt to scene changes at different angles, thereby achieving more accurate global positioning in complex environments.

[0027] 3) Local feature encoding generates global features to improve feature robustness: The present invention encodes local features to generate global features, and further optimizes feature expression by rotating data enhancement. Compared with the prior art, the global features generated in this way are more robust, can better adapt to environmental changes and noise interference, and show higher robustness and accuracy.

[0028] 4) Local feature brute force matching to improve computational efficiency: In the local feature matching stage, the present invention adopts a brute force matching method to calculate the relative transformation matrix. On the basis of ensuring feature accuracy, the brute force matching method can effectively improve the calculation efficiency and reduce the time overhead of global relocation, thereby meeting the real-time requirements of the mobile robot.

[0029] 5) Easy to deploy and reduce resource requirements: The present invention significantly reduces the consumption of computing resources by optimizing point cloud data processing and feature extraction without introducing additional sensors. At the same time, the method is easy to deploy into the existing mobile robot system without complicated hardware modification, and has high practicality and promotion value.

[0030] In summary, the present invention realizes the high-precision, robust, efficient and easy-to-deploy global repositioning of the mobile robot in the global map by using key technologies such as point cloud data generation of BEV images, rotation data enhancement, local feature encoding generation of global features, and brute force matching, and improves the global positioning ability of the mobile robot in the case of lost positioning or no initialization posture, thereby improving its reliability and efficiency in practical applications. In addition, the method shows excellent performance in complex environments, can effectively solve many problems existing in the prior art, and provides strong technical support for the autonomous navigation and global repositioning of mobile robots. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a principle block diagram when the method of the present invention is implemented; Figure 2 It is a BEV picture obtained in the mapping stage when the method of the present invention is implemented; Figure 3It is a flow chart of the positioning stage when implementing the method of the present invention; Figure 4 A schematic diagram of corresponding points obtained during the positioning phase of the method of the present invention; Figure 5 It is a principle block diagram when the system of the present invention is implemented; Figure 6 It is a principle block diagram when the device of the present invention is implemented. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be described in detail below in conjunction with specific embodiments. It should be clear that the following embodiments are only used to explain the present invention, but not to limit the scope of the present invention. Based on the technical scheme of the present invention, those of ordinary skill in the art can derive other feasible implementation methods without paying creative work, and these implementation methods all belong to the protection scope of the present invention.

[0033] Furthermore, the embodiments of the present invention are intended to help those skilled in the art better understand the technical solutions of the present invention through specific technical details and implementations. However, it should be understood by those skilled in the art that the implementation of the present invention is not limited to the specific details described in the following embodiments. Without departing from the core idea of ​​the present invention, the technical details in the embodiments may be appropriately adjusted, replaced or combined to adapt to different application scenarios or requirements.

[0034] It should be noted that the present invention is applicable to a mobile robot carrying a laser radar, an IMU and a wheel speed meter. The mobile robot obtains its own position information in the global map in real time through the SLAM method, thereby realizing the function of performing a specific task or completing a specific operation at a specific location.

[0035] Furthermore, the present invention is applied to the positioning task of a mobile robot in a global map. When the robot cannot obtain its position information in the global map due to some reasons, the technical solution proposed by the present invention can be used to obtain accurate global position information of the robot.

[0036] Embodiment 1: The global repositioning method of the mobile robot of the present invention.

[0037] like Figure 1 As shown, this embodiment provides a global relocalization method for a mobile robot, including a mapping phase and a localization phase, which are specifically described as follows.

[0038] S1. Before the robot performs a task or operation, it first enters the mapping phase. Figure 1 As shown, the mapping stage includes the following steps: S1-1. The robot takes the starting position of the mapping operation as the map origin, and during the movement, estimates the movement distance of the robot relative to the starting position through the odometer, and at the same time, updates the robot's position information in real time.

[0039] In specific implementation, the robot pose usually includes position (x, y, z) and attitude (such as Euler angles or quaternions). The relative position and attitude change of the robot in the global coordinate system are estimated by fusing sensor data such as wheel speed meter and IMU (inertial measurement unit). In addition, the pose information is usually represented in matrix form, for example: ‌pose matrix‌: a 4x4 transformation matrix containing rotation and translation information; ‌pose vector‌: a 6-dimensional vector containing position (x, y, z) and attitude (roll, pitch, yaw).

[0040] S1-2, during the movement of the robot, the laser radar scans and obtains the scan frame, thereby continuously acquiring the surrounding environment information. Whenever the robot moves a fixed distance (such as 0.5 meters), the current laser radar scan frame is used as the key frame.

[0041] S1-3. For each key frame, obtain the current robot's position and posture information from the odometer and record it as the robot's position and posture of the key frame.

[0042] In specific implementation, the posture recording and key frame generation are synchronized. When the robot moves 0.5 meters and generates a key frame, the system will also record the current posture information of the robot.

[0043] S1-4, for each key frame point cloud data, generate a BEV image by voxelization, such as Figure 2 shown.

[0044] In specific implementation, the point cloud data is converted into BEV images through voxelization, thereby reducing the impact of point cloud sparsity on feature extraction.

[0045] S1-5. Perform rotation data enhancement on each BEV image, rotating 45 degrees each time, and generate 8 rotated BEV images.

[0046] In the specific implementation, the input is augmented with rotation changes, that is, the BEV image is rotated 45 degrees each time. Through this data augmentation method, each BEV image can eventually become 8 images. Correspondingly, when encoding local features into global features, the network input data becomes eight times the original, thereby increasing the diversity of the input data of the subsequent global feature extraction network model and enhancing the network's robustness and generalization ability to scene data rotation changes.

[0047] In addition, the data enhancement in this embodiment is achieved by rotating the input data 45 degrees each time, and other rotation angles or data enhancement methods can also be used. Moreover, the angle and number of rotation data enhancement can be adjusted according to the rotation change characteristics of the scene to further improve the robustness and generalization ability of the network.

[0048] S1-6. Extract local features of each BEV image through the residual network Resnet.

[0049] In the specific implementation, this embodiment selects the residual network Resnet to extract the local features of the BEV image. In addition to the residual network Resnet, the method of the present invention can also select other methods to extract the local features of the BEV image, such as SIFT, SURF, ORB, SuperPoint, SuperGlue, etc.

[0050] S1-7. Input the local features of each BEV image into the NetVLAD network and encode them to generate robust global features.

[0051] In the specific implementation, this embodiment selects the NetVLAD network to encode local features to generate robust global features, and the positioning accuracy is improved through the residual network Resnet and the NetVLAD network. In addition to the NetVLAD network, the present invention can also select other methods to generate global features, such as Patch-NetVLAD, Super-NetVLAD, etc.

[0052] S1-8, bind the robot posture, BEV image, local features and global features corresponding to each key frame, and store them in the feature library.

[0053] In specific implementation, when each keyframe is generated, the system will record the current robot's posture information and store it in conjunction with other keyframe data (such as BEV images, local features, and global features). This association provides the necessary information for subsequent global relocalization.

[0054] S1-9. After the map is built, the point cloud data of the global map and the feature library consisting of all key frames are generated.

[0055] In specific implementation, the feature library includes the robot posture, BEV images, local features and global features corresponding to the key frames, which are used for robot retrieval and positioning without a global position.

[0056] S2: When the robot has no global position, such as lost positioning or no initialization posture, global relocalization will be triggered. Figure 1 and Figure 3 As shown, the positioning stage includes the following steps.

[0057] S2-1: During the positioning phase, the robot obtains the point cloud data of the current scanning frame through lidar scanning.

[0058] S2-2, generating a BEV image by voxelizing the point cloud data of the current scan frame.

[0059] S2-3. Extract the local features of the BEV image through the residual network Resnet, and encode the local features to generate global features through the NetVLAD network.

[0060] In specific implementation, in addition to the residual network Resnet selected in this embodiment, the method of the present invention also selects other methods to extract local features of BEV images, such as SIFT, SURF, ORB, SuperPoint, SuperGlue, etc.

[0061] Furthermore, in addition to the NetVLAD network selected in this embodiment, the present invention may also select other methods to generate global features, such as Patch-NetVLAD, Super-NetVLAD, etc.

[0062] S2-4. Retrieve the global features that match the global features of the current scanning frame in the feature library, and obtain the local features and robot posture information corresponding to the matched global features.

[0063] In the specific implementation, the feature library data generated in the mapping stage is loaded, and the global features most similar to the global features of the current scanning frame are retrieved in the feature library to obtain the corresponding local features and robot posture information.

[0064] S2-5, the local features of the current scan frame are matched with the retrieved local features by brute force to obtain corresponding points, such as Figure 4 shown.

[0065] In the specific implementation, the brute force matching in this embodiment adopts a feature point matching algorithm. In addition, the specific algorithm of local feature encoding and matching can be selected or optimized according to computing resources and real-time requirements, as long as it can achieve efficient and accurate global relocation.

[0066] S2-6. The corresponding points are estimated by the RANSAC algorithm to obtain the relative transformation matrix between the local features of the current scanning frame and the acquired local features.

[0067] In the specific implementation, this step selects the RANSAC algorithm to estimate the relative transformation matrix between two local features. In addition, efficient and accurate global relocalization is achieved through brute force matching and the RANSAC algorithm.

[0068] S2-7. Apply the relative transformation matrix to the retrieved robot posture to obtain the current global position information of the mobile robot.

[0069] As can be seen from the above, the method of the present invention is applicable to the positioning task of a mobile robot equipped with a laser radar, an IMU and a wheel speed meter in a global map. In particular, the global repositioning of the mobile robot is achieved without introducing additional sensors. By aligning the point cloud data obtained by each scan of the laser radar with the point cloud of the global map, the real-time position and posture of the robot in the global map are obtained, and a more accurate, robust, efficient and easy-to-deploy global positioning of the mobile robot is achieved, and the global positioning capability of the robot is improved in the case of lost positioning or no initialization posture. It can be widely used in industrial automation, warehousing and logistics, service robots and other fields.

[0070] Embodiment 2: The global repositioning system of the mobile robot of the present invention.

[0071] like Figure 5 As shown, this embodiment provides a mobile robot global repositioning system, which utilizes a mobile robot global repositioning method described in Example 1, including a mapping module and a positioning module, which are specifically introduced as follows.

[0072] (1) The mapping module includes a key frame generation unit, a key frame processing unit, and a feature library generation unit.

[0073] Among them, the key frame generation unit is used for: the robot takes the starting position of the mapping operation as the map origin, estimates the movement distance of the robot relative to the starting position through the odometer, and updates the robot's posture information in real time; every time the robot moves a fixed distance, the scanning frame obtained by the current laser radar is used as the key frame, and the current robot's posture information is obtained from the odometer as the robot posture of the key frame.

[0074] The key frame processing unit is used to: generate a BEV picture by voxelization for the point cloud data of each key frame; then perform rotation data enhancement on each BEV picture to generate multiple rotated BEV pictures; then extract local features of each BEV picture and generate global features from the local features.

[0075] The feature library generation unit is used to: bind the robot posture, BEV image, local features and global features corresponding to each key frame, and store them in the feature library; after the map is built, the point cloud data of the global map and the feature library composed of all key frames are obtained.

[0076] (2) The positioning module includes a global feature generation unit, a feature library retrieval unit, and a global position acquisition unit.

[0077] The global feature generation unit is used for: the robot obtains the current scanning frame through laser radar scanning; generates a BEV picture by voxelization for the point cloud data of the current scanning frame; then extracts the local features of each BEV picture, and generates global features from the local features.

[0078] The feature library retrieval unit is used to retrieve global features that match the global features of the current scanning frame in the feature library, and obtain local features and robot posture information corresponding to the matched global features.

[0079] The global position acquisition unit is used to: obtain corresponding points by brute force matching the local features of the current scanning frame with the acquired local features; and estimate the posture of the corresponding points to obtain the relative transformation matrix between the local features of the current scanning frame and the acquired local features; and then apply the relative transformation matrix to the acquired robot posture information to obtain the current global position information of the robot.

[0080] In summary, through the above technical solutions, the system of the present invention realizes the accurate mapping of scenes by mobile robots, as well as fast and accurate positioning in the mapped scenes, thereby improving the global positioning capability of the robot in the case of lost positioning or no initialized posture, thereby improving its reliability and efficiency in practical applications.

[0081] Embodiment 3: The global repositioning device of the mobile robot of the present invention.

[0082] like Figure 6 As shown, this embodiment provides a global repositioning device for a mobile robot, comprising at least one processor; and a memory connected to at least one of the processors.

[0083] The memory stores instructions that can be executed by the processor.

[0084] The instructions are used to be executed by the processor to implement a mobile robot global repositioning method described in Example 1.

[0085] In a specific implementation, the processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the device to perform desired functions and improve algorithm efficiency.

[0086] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the instructions to implement the methods of the various embodiments described above and / or the functions of the system thereof.

[0087] The instructions are computer program codes written in one or more programming languages ​​or a combination thereof for performing the operations of the present application, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server.

[0088] Embodiment 4: Computer readable storage medium of the present invention.

[0089] This embodiment provides a computer-readable storage medium having a computer program stored thereon; when the computer program is executed by a processor, the mobile robot global repositioning method described in Embodiment 1 is implemented.

[0090] In specific implementation, the computer program can be written in any combination of one or more programming languages ​​to execute the program code of the embodiment of the present application, and the programming language includes an object-oriented programming language, such as Java, C++, etc., and also includes a conventional procedural programming language, such as "C" language or similar programming languages. The program code can be executed completely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on the remote computing device, or completely on the remote computing device or server.

[0091] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0092] In addition, the block diagrams of the systems and devices to which the present invention relates are only illustrative examples, and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these systems and devices may be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open words, meaning "including but not limited to," and may be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or," and may be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and may be used interchangeably therewith.

[0093] Furthermore, it should be understood by those skilled in the art that the above modules or steps of the present invention can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

Claims

1. A global relocalization method for a mobile robot, characterized in that: Includes mapping phase and positioning phase; The mapping stage includes the following steps: The robot uses the starting position of the mapping operation as the map origin, estimates the movement distance of the robot relative to the starting position through the odometer, and updates the robot's posture information in real time; every time the robot moves a fixed distance, the scan frame obtained by the current lidar is used as the key frame, and the current robot posture information obtained from the odometer is used as the robot posture of the key frame; For the point cloud data of each key frame, a BEV image is generated by voxelization; then, each BEV image is rotated to enhance the data and generate multiple rotated BEV images; then, the local features of each BEV image are extracted and encoded to generate global features; The robot posture, BEV image, local features and global features corresponding to each key frame are bound and stored in the feature library; after the map is built, the point cloud data of the global map and the feature library composed of all key frames are obtained; The positioning stage includes the following steps: The robot obtains the current scan frame through laser radar scanning; generates a BEV image by voxelizing the point cloud data of the current scan frame; then extracts the local features of each BEV image, and encodes the local features to generate global features; Retrieve the global features that match the global features of the current scan frame in the feature library, and obtain the local features and robot posture information corresponding to the matched global features; The local features of the current scanning frame are matched with the acquired local features by brute force to obtain corresponding points; the corresponding points are estimated to obtain the relative transformation matrix between the local features of the current scanning frame and the acquired local features; the relative transformation matrix is ​​then applied to the acquired robot posture information to obtain the robot's current global position information.

2. A mobile robot global relocation method according to claim 1, characterized in that: The fixed distance is 0.5 meters.

3. A mobile robot global relocation method according to claim 1, characterized in that: The rotation data enhancement method is to rotate 45 degrees each time to generate 8 rotated BEV pictures.

4. A mobile robot global relocation method according to claim 1, characterized in that: The local features are extracted through the residual network Resnet.

5. A mobile robot global relocation method according to claim 1 or 4, characterized in that: The global features are generated through NetVLAD network coding.

6. A mobile robot global relocation method according to claim 1, characterized in that: The brute force matching adopts a feature point matching algorithm.

7. A mobile robot global relocation method according to claim 1 or 6, characterized in that: The pose estimation adopts the RANSAC algorithm.

8. A mobile robot global repositioning system, using a mobile robot global repositioning method as claimed in claim 1, characterized in that: Includes mapping module and positioning module; The mapping module includes a key frame generation unit, a key frame processing unit, and a feature library generation unit; The key frame generation unit is used to: the robot takes the starting position of the map construction operation as the map origin, estimates the movement distance of the robot relative to the starting position through the odometer, and updates the robot's posture information in real time; each time the robot moves a fixed distance, the scanning frame obtained by the current laser radar is used as the key frame, and the current robot posture information is obtained from the odometer as the robot posture of the key frame; The key frame processing unit is used to: generate a BEV image by voxelization for the point cloud data of each key frame; then perform rotation data enhancement on each BEV image to generate multiple rotated BEV images; then extract local features of each BEV image, and generate global features from the local features; The feature library generation unit is used to: bind the robot posture, BEV image, local features and global features corresponding to each key frame, and store them in the feature library; after the map is built, the point cloud data of the global map and the feature library composed of all key frames are obtained; The positioning module includes a global feature generation unit, a feature library retrieval unit, and a global position acquisition unit; The global feature generation unit is used to: the robot obtains the current scanning frame through laser radar scanning; generates a BEV picture by voxelization for the point cloud data of the current scanning frame; then extracts the local features of each BEV picture, and generates global features from the local features; The feature library retrieval unit is used to: retrieve the global features matching the global features of the current scanning frame in the feature library, and obtain the local features and robot posture information corresponding to the matching global features; The global position acquisition unit is used to: obtain corresponding points by brute force matching the local features of the current scanning frame with the acquired local features; and estimate the posture of the corresponding points to obtain the relative transformation matrix between the local features of the current scanning frame and the acquired local features; and then apply the relative transformation matrix to the acquired robot posture information to obtain the current global position information of the robot.

9. A global repositioning device for a mobile robot, characterized in that: comprising at least one processor; and a memory connected to at least one of the processors; Wherein, the memory stores instructions executable by the processor; The instructions are used to be executed by the processor to implement a mobile robot global repositioning method as described in any one of claims 2 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, a mobile robot global repositioning method as described in any one of claims 2 to 7 is implemented.

Images