Full-autonomous navigation method and device for indoor sparse feature mobile robot

Abstract

The invention provides an indoor sparse feature mobile robot full-autonomous navigation method and device, and relates to the technical field of mobile robots. The objective of the invention is to solve the problems that in a long corridor of a hospital, laser radar point clouds are sparsely collected, the AMR position cannot be accurately positioned, positioning is wrong or a robot is bound, an improved algorithm is provided for trajectory tracking, two-dimensional laser radar point cloud features are combined with a classifier based on a deep convolutional network, and the positioning accuracy is improved. The AMR can determine the position of the robot to distinguish the situation where the SLAM or ranging positioning system is currently selected, i.e., the task can still continue if the AMR is in the situation where the position of the robot cannot be determined.

Description

Technical Field

[0001] The present invention relates to the field of mobile robots, and in particular to a fully autonomous navigation method and device for an indoor sparse feature mobile robot. Background Art

[0002] In recent years, due to the rapid development and development of technology, more and more industries are turning to automation. Autonomous mobile robots (AMRs) play an important role in the automation process and are widely used in various fields. The robot "kidnapping" problem refers to determining the current position of the robot without its previous position information, such as when the robot is docked in a well-mapped environment, but its relative position in the map is unknown, or during the movement, due to temporary functional failure of the sensor or rapid movement of the camera, the robot's previous position information is lost.

[0003] Path planning is crucial for finding the best path from one point to another in an environment. The two mainstream methods are grid search-based methods and intelligence-based methods. Grid search-based methods include the A* algorithm and its variants. Although this method will produce some accuracy loss, the planned path is conservative. In order to reduce the redundant points in the path-finding process of the A* algorithm, jump points are obtained in the grid map through jump point search, and the speed of the A* algorithm is accelerated based on the obtained jump points, but the search time fluctuates greatly in different actual scenarios. The intelligence-based method usually improves the genetic algorithm under the global static environment to improve the slow convergence speed and precoding problems. In addition, particle swarm optimization is used to refine the inertia weight and acceleration factor to prevent the local minimum from falling and improve the convergence speed. In the path planning process, trajectory tracking is required so that the AMR can track the movement according to the set trajectory path. The most common method is to use the local linear model predictive control (MPC) method to simultaneously track the speed and path of the nonlinear robot car model. When obstacles appear in the predicted horizon, the trajectory is dynamically adjusted through the nonlinear MPC. In addition, the adaptive trajectory tracking strategy and the hybrid vision trajectory strategy can optimize the controller parameter settings according to different combinations of vehicle speed and heading errors. The above algorithms are not sufficient to support fully autonomous navigation of robots under sparse feature conditions indoors. Their limitations are computational difficulties and the inability to provide a closed-loop solution when the model is complex. In addition, traditional trajectory tracking algorithms (such as PP) do not properly select the forward-looking distance under limited resource conditions, resulting in poor tracking performance.

[0004] In order to navigate autonomously and safely, the AMR needs to locate its position in the scene of sparse features in the hospital. A simple form of positioning is to use odometry, which provides the current position through the estimated odometry information such as the speed and rotation of the wheel (wheel odometry), inertial measurement unit (imu odometry), laser source (laser odometry) or image (visual odometry). The first is to estimate the odometry using only the lidar odometry algorithm, and then segment the local map through the convolutional neural network (CNN), and then use two-stage RANSAC to verify the position matching in the local map. The second is the simultaneous localization and mapping (SLAM) technology, which includes a map building process and a localization process, generally based on least squares geometric matching to compensate the predicted position. Although the radar-based SLAM method provides useful information for determining the characteristics of free space areas and positioning positions, it seems to be inefficient in unstructured environments, such as long corridors, tunnels, dusty or foggy areas. Therefore, the present application proposes a fully autonomous navigation method and device for indoor sparse feature mobile robots to at least partially solve the problems that may exist in the prior art. Summary of the invention

[0005] In order to overcome the above problems or at least partially solve the above problems, an embodiment of the present invention provides a fully autonomous navigation method and device for an indoor sparse feature mobile robot, which combines two-dimensional lidar point cloud features with a classifier based on a deep convolutional network to distinguish the current selection of SLAM or ranging and positioning system, that is, if the AMR is in a situation where the robot position cannot be determined, the task can still continue.

[0006] The embodiment of the present invention is achieved as follows: The present application provides a method for realizing fully autonomous navigation of an indoor sparse feature mobile robot, the method comprising: Trajectory tracking: using the angle a between two vectors created from the current robot center to the next forward-looking point and the current rotation angle b to predict turning distance and perform steering actions; 2D laser SLAM: Based on the unstructured environment scene of a long corridor, the SLAM algorithm is used to match the point cloud with the map features through the two-dimensional laser radar data. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it can be determined whether the AMR is positioned incorrectly. Corridor recognition of the exchange positioning system based on deep learning: the point cloud data is converted into a two-dimensional image through a preset algorithm, and the point cloud information is drawn on the two-dimensional image, the obtained two-dimensional point cloud image is used as input for model training, and the deep learning method is deployed in the AGV equipment, wherein, when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the IMU-based odometer positioning system.

[0007] In some embodiments of the present invention, the trajectory tracking step specifically includes: Predict the turning distance using the angle a between two vectors created from the current robot center to the next look-ahead point; The current rotation angle between the two vectors created from the current robot center to the next and previous look-ahead points is defined as b, and ensure that b ≤ value, used for steering action; where, ideally, =135°, when the robot is driving on rough terrain, =0.1 (rad / s) is used to indicate angular velocity In greater than The threshold value when .

[0008] In some embodiments of the present invention, the predicting the turning distance using the angle a between two vectors created from two points from the current robot center to the next forward-looking point comprises: Use two vectors created from the current robot center to the two points of the next look-ahead point and , where the vector Used to predict distance, vector Defined as ; According to the vector and , determine the angle a between the two vectors, where a≤ Value; from and You can get The maximum value of , and the ideal angle a is 60°.

[0009] In some embodiments of the present invention, the steps of 2D laser SLAM include: The unstructured environment scenario based on a long corridor is to identify the lost position of the mobile robot through the features of the 2D LiDAR data; Match the point cloud with map features through the SLAM algorithm; If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and all point clouds and the point cloud superimposed on the black edge of the map are used to determine whether the AMR is positioned incorrectly.

[0010] In some embodiments of the present invention, if the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and all point clouds and the point cloud superimposed on the black edge of the map are used to determine whether the AMR is positioned incorrectly, including: Using all point clouds and the point cloud superimposed on the black edge of the map, determine the missing rate Value, among which, missing rate , which ranges from 0 to 1; represents the number of all point clouds in a frame, and Represents the number of point clouds in a frame that are superimposed on map features; If it is greater than 50%, most of the point clouds will not be superimposed on the map features, thus determining that the mobile robot is lost, otherwise, it indicates that the positioning is successful; or, when the surrounding environment scene has a repetitive pattern and the point cloud is superimposed on the map features, the map features will be too consistent, making it impossible to confirm the SLAM localization and determine the position of the mobile robot in the corridor.

[0011] In some embodiments of the present invention, the corridor identification of the exchange positioning system based on deep learning specifically includes: Through the following formula, Convert the LiDAR point cloud data into a two-dimensional image, where is the position of the point cloud on the image, is the location of the point cloud in the real world, is the transfer matrix from the lidar point cloud position to the image point cloud position, is the offset between the lidar point cloud position and the image point cloud position; By setting the following parameters, one pixel is set to be equal to 0.05m. For the image resolution, the point cloud range is set within a square range of 10m × 10m, with the center of the mobile robot as the base point, and the real scale is converted into image pixels; Through the transformation matrix, the map coordinates (100, 200) are used as the center of the mobile robot to draw the point cloud information on the two-dimensional image; Image edge detection is used to determine the region of interest (ROI) within the image range of x ≥ 100 and 90 ≤ y ≤ 110, and the content of the region of interest is filtered to obtain a noise-free image; The noise-free image is fed into a deep neural network to determine the corridor area. The improved Inception V3 deep neural network architecture is used as the corridor recognition network. In the corridor classification task, the fine-tuning method is used based on the ImageNet dataset to deploy the deep learning method to the AGV equipment. When the trained deep neural network detects a long corridor area, the AGV switches from the SLAM positioning system to the IMU-based odometry positioning system.

[0012] An indoor sparse feature mobile robot fully autonomous navigation device, comprising: Trajectory tracking module for predicting turning distance and performing steering actions using the angle a between two vectors created from two points from the current robot center to the next forward-looking point and the current rotation angle b; 2D laser SLAM module, used in unstructured environment scenarios based on long corridors, uses 2D laser radar data and SLAM algorithm to match point clouds with map features. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it can be determined whether the AMR is positioned incorrectly. The system positioning module is used to convert point cloud data into a two-dimensional image through a preset algorithm, draw point cloud information on the two-dimensional image, use the obtained two-dimensional point cloud image as input for model training, and deploy the deep learning method to the AGV equipment, wherein when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the IMU-based odometer positioning system.

[0013] An electronic device comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the method for realizing fully autonomous navigation of an indoor sparse feature mobile robot is implemented.

[0014] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for realizing fully autonomous navigation of an indoor sparse feature mobile robot is implemented.

[0015] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: The angle a between two vectors created by using two points from the current robot center to the next forward-looking point and the current rotation angle b are used to predict the turning distance and perform steering actions; based on the unstructured environment scene of the long corridor, the point cloud is matched with the map features through the two-dimensional lidar data using the SLAM algorithm, wherein if the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it is determined whether the AMR is positioned incorrectly; the point cloud data is converted into a two-dimensional image through a preset algorithm, and the point cloud information is drawn on the two-dimensional image, the obtained two-dimensional point cloud image is used as input for model training, and the deep learning method is deployed in the AGV device, wherein when the trained deep neural network detects the long corridor area, the AGV switches the SLAM positioning system to the imu-based odometer positioning system. The purpose of the present invention is to solve the problem that the AMR position cannot be accurately located due to sparse collection of LiDAR point clouds in long hospital corridors, as well as positioning errors or robot abduction, and an improved algorithm is provided for trajectory tracking, which combines the 2D LiDAR point cloud features with a classifier based on a deep convolutional network to distinguish between the current selection of SLAM or ranging positioning system, that is, if the AMR is in a situation where the robot position cannot be determined, the task can still be continued. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A flowchart of the steps of a fully autonomous navigation method for an indoor sparse feature mobile robot provided by an embodiment of the present invention; Figure 2 A schematic diagram of a coordinate system of a two-wheel differential mobile robot kinematic model for a fully autonomous navigation method of an indoor sparse feature mobile robot provided by an embodiment of the present invention; Figure 3 A schematic diagram of a turning decision strategy during robot trajectory tracking in a fully autonomous navigation method for an indoor sparse feature mobile robot provided by an embodiment of the present invention; Figure 4 A schematic diagram of a mobile robot hardware platform for a fully autonomous navigation method of an indoor sparse feature mobile robot provided by an embodiment of the present invention; Figure 5 A schematic diagram of the module structure of a fully autonomous navigation device for an indoor sparse feature mobile robot provided by an embodiment of the present invention; Figure 6A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the objects, features and advantages of the present application more obvious and understandable, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0019] The core concept of the present invention is a fully autonomous navigation method for a sparse feature mobile robot in an indoor scene, focusing on the control ability of AMR on a curved path and the positioning problem of AMR in an unstructured environment (long corridor). For trajectory tracking, in order to solve the error problem caused by the traditional method of often slowing down when encountering a turn, the improved PP algorithm is combined with the deceleration based on the prediction of the turn to reduce the impact of delayed deceleration attempts. In order to solve the problem of the kidnapped robot, the two-dimensional lidar point cloud features are combined with a classifier based on a deep convolutional network to distinguish the current selection of SLAM or range measurement positioning system. Therefore, if the AMR is in a situation where SLAM cannot determine the position of the robot, the task can still continue.

[0020] Reference Figure 1 , showing a fully autonomous navigation method for an indoor sparse feature mobile robot provided by an embodiment of the present application; the method comprises: S100, a step of trajectory tracking: using an angle a between two vectors created from two points from the current robot center to the next forward-looking point and a current rotation angle b to predict a turning distance and perform a turning action; S200, 2D laser SLAM step: Based on the unstructured environment scene of the long corridor, the SLAM algorithm is used to match the point cloud with the map features through the two-dimensional laser radar data. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it is determined whether the AMR is positioned incorrectly; S300, corridor recognition step of the exchange positioning system based on deep learning: converting point cloud data into a two-dimensional image through a preset algorithm, drawing point cloud information on the two-dimensional image, using the obtained two-dimensional point cloud image as input for model training, and deploying the deep learning method to the AGV device, wherein when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the IMU-based odometer positioning system.

[0021] The purpose of the present invention is to solve the problem that the AMR position cannot be accurately located due to sparse collection of LiDAR point clouds in long hospital corridors, as well as positioning errors or robot abduction, and an improved algorithm is provided for trajectory tracking, which combines the 2D LiDAR point cloud features with a classifier based on a deep convolutional network to distinguish between the current selection of SLAM or ranging positioning system, that is, if the AMR is in a situation where the robot position cannot be determined, the task can still be continued.

[0022] It should be noted that IMU (Inertial Measurement Unit): the core sensor for attitude measurement and inertial navigation. IMU is a device for measuring the three-axis angular rate and acceleration of an object, usually composed of two or more accelerometers and gyroscopes. As the core equipment of inertial positioning technology, IMU finally outputs navigation information such as the coordinate change and speed of the carrier relative to the initial position after error compensation and inertial navigation calculation. IMU can be divided into consumer, industrial and military grades according to its performance from low to high. Different levels of IMU differ in design, manufacturing and cost to meet the performance and reliability requirements of different fields. IMU has a wide range of downstream applications, covering various fields from consumer electronics, automotive electronics to industry and communications, medical health, etc.

[0023] Next, the fully autonomous navigation method of the indoor sparse feature mobile robot in this exemplary embodiment will be further described.

[0024] In the above step S100, the trajectory tracking step: using the angle a between two vectors created from two points from the current robot center to the next forward-looking point and the current rotation angle b to predict the turning distance and perform the turning action; including: Predict the turning distance using the angle a between two vectors created from the current robot center to the next look-ahead point; The current rotation angle between the two vectors created from the current robot center to the next and previous look-ahead points is defined as b, and ensure that b ≤ value, used for steering action; where, ideally, =135°, when the robot is driving on rough terrain, =0.1 (rad / s) is used to indicate angular velocity In greater than The threshold value when , helps the mobile robot to improve the tracking accuracy.

[0025] Further, the method predicts the turning distance using the angle a between two vectors created from two points from the current robot center to the next forward-looking point, including: Use two vectors created from the current robot center to the two points of the next look-ahead point and , where the vector Used to predict distance, vector Defined as ; According to the vector and , determine the angle a between the two vectors, where a≤ Value; from and You can get The maximum value of , and the ideal angle a is 60°.

[0026] As an example, the mobile robot of the present invention is driven by a differential wheel mode. The left and right wheels are related to the overall speed and angular velocity of the mobile robot. The coordinate system of the mobile robot is as follows: Figure 2 As shown, is the position of the mobile robot, , , L is the distance between the left wheels, is the angle between the mobile robot and the x-axis, is the speed of the right wheel, is the speed of the left wheel, is the speed of the mobile robot, is the angular velocity of the mobile robot.

[0027] When the angular velocity ω is large, the turning time can be predicted. However, the long lookahead distance reduces the accuracy of trajectory tracking, making it impossible to slow down the turning speed over time. To address the above problem, a fixed short lookahead distance is used to improve the trajectory tracking capability anytime and anywhere. Then, the method determines the deceleration speed of the current turn and converts the angular velocity Keep it level so that the mobile robot can follow the path better when the path becomes more rugged. Figure 3 As shown, the present invention divides the steering decision strategy into three steps: First, the turning distance is predicted using the angle a between two vectors created from the current robot center to the next forward-looking point. Assume that in an ideal case, the two vectors at angle a (the ideal angle is 60◦) (prediction distance), Defined as When the ideal situation is encountered, the AMR will slow down. However, when the turning radius of the route is R, the angle a will never reach value, AMR cannot decelerate. Without considering R, and You can get The maximum value of .

[0028] Second, to perform the steering action, the current rotation angle between the two vectors created from the current robot center to the next and previous look-ahead points is defined as b, and ensure that b ≤ In ideal conditions, =135◦. When the robot is driving on rough terrain, =0.1 (rad / s) is used to indicate angular velocity In greater than The threshold value when , helps the mobile robot to improve the tracking accuracy.

[0029] In one embodiment of the present invention, the specific process of “2D laser SLAM” described in step S200 may be further explained in combination with the following description.

[0030] Based on the unstructured environment scene of a long corridor, the SLAM algorithm is used to match the point cloud with the map features through the two-dimensional lidar data. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it can be determined whether the AMR is positioned incorrectly; including: The unstructured environment scenario based on a long corridor is to identify the lost position of the mobile robot through the features of the 2D LiDAR data; Match the point cloud with map features through the SLAM algorithm; If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and all point clouds and the point cloud superimposed on the black edge of the map are used to determine whether the AMR is positioned incorrectly.

[0031] Further, the point cloud will be superimposed on the black edge of the SLAM map, and all point clouds and the point cloud superimposed on the black edge of the map will be used to determine whether the AMR is mispositioned, including: Determine the missing rate using all point clouds and the point cloud superimposed on the black edge of the map Value, among which, missing rate , which ranges from 0 to 1; represents the number of all point clouds in a frame, and Represents the number of point clouds in a frame that are superimposed on map features; If it is greater than 50%, most of the point clouds will not be superimposed on the map features, thus determining that the mobile robot is lost, otherwise, it indicates that the positioning is successful; or, when the surrounding environment scene has a repetitive pattern and the point cloud is superimposed on the map features, the map features will be too consistent, making it impossible to confirm the SLAM localization and determine the position of the mobile robot in the corridor.

[0032] As an example, in the "2D laser SLAM" described in step S200, in the unstructured environment of a long corridor, SLAM often cannot determine its position on the map, which can easily cause accidents. To avoid positioning errors, the present invention identifies the lost position of the mobile robot based on the characteristics of the two-dimensional lidar data. First, in scan matching, SLAM matches the point cloud with the map features. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map. At this time, all point clouds and the point cloud superimposed on the black edge of the map are used to determine whether the AMR is mispositioned. Define the missing rate (ranges from 0 to 1), if If it is greater than 50%, most of the point clouds will not be superimposed on the map features. In this case, it can be judged that the mobile robot is losing; otherwise, it means that the positioning is successful. represents the number of all point clouds in a frame, and Represents the number of point clouds in one frame that are superimposed on map features. When the surrounding environment has a repetitive pattern, although the point clouds are superimposed on the map features, the map features will be too consistent to make SLAM localization confirmed. The point clouds extracted from the walls on both sides are too sparse to determine the position of the mobile robot in the corridor, which will be solved in the next step.

[0033] In one embodiment of the present invention, the specific process of the "corridor identification step of the exchange positioning system based on deep learning" described in step S300 can be further explained in combination with the following description.

[0034] The point cloud data is converted into a two-dimensional image by a preset algorithm, and the point cloud information is drawn on the two-dimensional image, the obtained two-dimensional point cloud image is used as input for model training, and the deep learning method is deployed in the AGV device, wherein, when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the odometer positioning system based on the imu, specifically including: converting the point cloud data into a two-dimensional image by a preset algorithm, and drawing the point cloud information on the two-dimensional image, the obtained two-dimensional point cloud image is used as input for model training, and the deep learning method is deployed in the AGV device, wherein, when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the odometer positioning system based on the imu; Image collection: Through the following formula, Convert the LiDAR point cloud data into a two-dimensional image, where is the position of the point cloud on the image, is the location of the point cloud in the real world, is the transfer matrix from the lidar point cloud position to the image point cloud position, is the offset between the laser radar point cloud position and the image point cloud position; refer to Figure 4 The figure shows a schematic diagram of the mobile robot hardware platform. The depth image and lidar data are obtained through the depth camera and 2D lidar integrated in the mobile robot hardware platform.

[0035] By setting the following parameters, one pixel is set to be equal to 0.05m. For the image resolution, the point cloud range is set within a square range of 10m × 10m, with the center of the mobile robot as the base point, and the real scale is converted into image pixels; Through the transformation matrix, the map coordinates (100, 200) are used as the center of the mobile robot to draw the point cloud information on the two-dimensional image; Image edge detection is used to determine the region of interest (ROI) within the image range of x ≥ 100 and 90 ≤ y ≤ 110, and the content of the region of interest is filtered to obtain a noise-free image; The noise-free image is fed into a deep neural network to determine the corridor area. The improved Inception V3 deep neural network architecture is used as the corridor recognition network. In the corridor classification task, the fine-tuning method is used based on the ImageNet dataset to deploy the deep learning method to the AGV equipment. When the trained deep neural network detects a long corridor area, the AGV switches from the SLAM positioning system to the IMU-based odometry positioning system.

[0036] As an example, the present invention uses deep learning to identify the location of the corridor area to switch the positioning system, and defines the corridor identification problem as a binary classification problem. The specific process is as follows: (1) First, in order to collect images representing the current area, we need to Convert the LiDAR point cloud data into a two-dimensional image. is the position of the point cloud on the image, is the location of the point cloud in the real world, is the transfer matrix from the lidar point cloud position to the image point cloud position, is the offset between the laser radar point cloud position and the image point cloud position. In order to convert the real scale into image pixels, one pixel is set to be equal to 0.05m. is the image resolution. The point cloud range is set within a 10m×10m square range, with the center of the mobile robot as the base point. Finally, through the transformation matrix, the map coordinates (100, 200) are used as the center of the mobile robot to draw the point cloud information on the two-dimensional image.

[0037] (2) When the 2D point cloud image is put into the deep neural network for recognition, noise will be generated if there are people in the image, and the recognition performance of the corridor is poor. Therefore, image edge detection is used to determine the region of interest (ROI) within the image range of x ≥ 100 and 90 ≤ y ≤ 110. The ROI content is then filtered for noise. After image preprocessing, it is put into the deep neural network to determine the corridor area.

[0038] (3) For the corridor recognition network, an improved Inception V3 deep neural network architecture is used. In the corridor classification task, the improved model uses a fine-tuning method based on the ImageNet dataset to speed up training and improve model accuracy. Its parameters are fewer than those of the Inception V3-based model, but it maintains good classification performance. Moreover, it does not require powerful hardware support to achieve its heavy calculations, and it is easy to deploy deep learning methods to AGV equipment.

[0039] (4) When the trained deep neural network detects a long corridor area, the AGV avoids the mislocalization problem by switching the SLAM positioning system to the IMU-based odometer positioning system.

[0040] like Figure 4 As shown, an indoor sparse feature mobile robot fully autonomous navigation device provided by an embodiment of the present application is shown, comprising: A trajectory tracking module 100 for predicting a turning distance and performing a turning action using an angle a between two vectors created from two points from the current robot center to the next forward-looking point and a current rotation angle b; 2D laser SLAM module 200, used in unstructured environment scenarios based on long corridors, uses 2D laser radar data and SLAM algorithm to match point clouds with map features. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it can be determined whether the AMR is positioned incorrectly; The system positioning module 300 is used to convert the point cloud data into a two-dimensional image through a preset algorithm, draw the point cloud information on the two-dimensional image, use the obtained two-dimensional point cloud image as input for model training, and deploy the deep learning method to the AGV device, wherein when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the IMU-based odometer positioning system.

[0041] Reference Figure 6 , showing a computer device of a fully autonomous navigation method of an indoor sparse feature mobile robot of the present invention, which may specifically include the following: The computer device 12 is in the form of a general-purpose computing device, and the components of the computer device 12 may include but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0042] The bus 18 represents one or more of several types of bus 18 structures, including a memory bus 18 or memory controller, a peripheral bus 18, an accelerated graphics port, a processor, or a local bus 18 using any of a variety of bus 18 architectures. These architectures include, by way of example, but are not limited to, an Industry Standard Architecture (ISA) bus 18, a Micro Channel Architecture (MAC) bus 18, an Enhanced ISA bus 18, an Audio Video Electronics Standards Association (VESA) local bus 18, and a Peripheral Component Interconnect (PCI) bus 18.

[0043] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0044] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). Although Figure 6 Not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM or other optical medium) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42, which are configured to perform the functions of various embodiments of the present invention.

[0045] A program / utility 40 having a set (at least one) of program modules 42 may be stored in, for example, a memory, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules 42, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0046] The computer device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, cameras, etc.), one or more devices that enable medical personnel to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the computer device 12 may also communicate with one or more networks (e.g., local area networks (LANs)), wide area networks (WANs), and / or public networks (e.g., the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the computer device 12 via a bus 18. It should be understood that although Figure 6 Not shown, other hardware and / or software modules may be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units 16, external disk drive arrays, RAID systems, tape drives, and data backup storage systems 34, etc.

[0047] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, such as implementing the fully autonomous navigation method of the indoor sparse feature mobile robot provided in the embodiment of the present invention.

[0048] That is, when the processing unit 16 executes the program, it achieves: using the angle a between two vectors created from two points from the current robot center to the next forward-looking point and the current rotation angle b to predict the turning distance and perform turning actions; based on the unstructured environment scene of the long corridor, through the two-dimensional lidar data, the point cloud is matched with the map features using the SLAM algorithm, wherein, if the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it is determined whether the AMR is positioned incorrectly; the point cloud data is converted into a two-dimensional image through a preset algorithm, and the point cloud information is drawn on the two-dimensional image, the obtained two-dimensional point cloud image is used as input for model training, and the deep learning method is deployed in the AGV device, wherein, when the trained deep neural network detects the long corridor area, the AGV switches the SLAM positioning system to the imu-based odometer positioning system.

[0049] In an embodiment of the present invention, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fully autonomous navigation method of an indoor sparse feature mobile robot provided in all embodiments of the present application: That is, when the program is executed by the processor, it is implemented as follows: the angle a between two vectors created from two points from the current robot center to the next forward-looking point and the current rotation angle b are used to predict the turning distance and perform turning actions; based on the unstructured environment scene of the long corridor, the point cloud is matched with the map features through the two-dimensional lidar data using the SLAM algorithm, wherein if the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it is determined whether the AMR is positioned incorrectly; the point cloud data is converted into a two-dimensional image through a preset algorithm, and the point cloud information is drawn on the two-dimensional image, the obtained two-dimensional point cloud image is used as input for model training, and the deep learning method is deployed in the AGV device, wherein when the trained deep neural network detects the long corridor area, the AGV switches the SLAM positioning system to the imu-based odometer positioning system.

[0050] Any combination of one or more computer-readable media may be used. A computer-readable medium may be a computer signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPOM 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 thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, device or device.

[0051] Computer-readable signal media may include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0052] The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the medical staff computer, partially on the medical staff computer, as an independent software package, partially on the medical staff computer, partially on the remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the medical staff computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same and similar parts between the various embodiments can be referred to each other.

[0053] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0054] Finally, 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 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 a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0055] The above is a detailed introduction to the fully autonomous navigation method and device for an indoor sparse feature mobile robot provided by the present application. This article uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, based on the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A fully autonomous navigation method for an indoor sparse feature mobile robot, characterized in that: include: Trajectory tracking: using the angle a between two vectors created from the current robot center to the next forward-looking point and the current rotation angle b to predict the turning distance and perform steering actions; 2D laser SLAM: Based on the unstructured environment scene of a long corridor, the SLAM algorithm is used to match the point cloud with the map features through the two-dimensional laser radar data. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it can be determined whether the AMR is positioned incorrectly. Corridor recognition of the exchange positioning system based on deep learning: the point cloud data is converted into a two-dimensional image through a preset algorithm, and the point cloud information is drawn on the two-dimensional image, the obtained two-dimensional point cloud image is used as input for model training, and the deep learning method is deployed in the AGV equipment, wherein, when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the IMU-based odometer positioning system.

2. The method according to claim 1, characterized in that The trajectory tracking step specifically includes: Predict the turning distance using the angle a between two vectors created from the current robot center to the next look-ahead point; The current rotation angle between the two vectors created from the current robot center to the next and previous look-ahead points is defined as b, and ensure that b ≤ value, used for steering action; where, ideally, =135°, when the robot is driving on rough terrain, =0.1 (rad / s) is used to indicate angular velocity In greater than The threshold value when .

3. The method according to claim 2, characterized in that The method predicts the turning distance using the angle a between two vectors created from two points from the current robot center to the next forward-looking point, including: Use two vectors created from the current robot center to the two points of the next look-ahead point and , where the vector Used to predict distance, vector Defined as ; According to the vector and , determine the angle a between the two vectors, where a≤ Value; from and You can get The maximum value of , and the ideal angle a is 60°.

4. The method according to claim 1, characterized in that: The steps of the 2D laser SLAM include: The unstructured environment scenario based on a long corridor is to identify the lost position of the mobile robot through the features of the 2D LiDAR data; Match the point cloud with map features through the SLAM algorithm; If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and all point clouds and the point cloud superimposed on the black edge of the map are used to determine whether the AMR is positioned incorrectly.

5. The method according to claim 4, characterized in that If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and all point clouds and the point cloud superimposed on the black edge of the map are used to determine whether the AMR is positioned incorrectly, including: Determine the missing rate using all point clouds and the point cloud superimposed on the black edge of the map Value, among which, missing rate , which ranges from 0 to 1; represents the number of all point clouds in a frame, and Represents the number of point clouds in a frame that are superimposed on map features; If it is greater than 50%, most of the point clouds will not be superimposed on the map features, thus determining that the mobile robot is lost, otherwise, it indicates that the positioning is successful; or, when the surrounding environment scene has a repetitive pattern and the point cloud is superimposed on the map features, the map features will be too consistent, making it impossible to confirm the SLAM localization and determine the position of the mobile robot in the corridor.

6. The method according to claim 1, characterized in that The corridor recognition of the exchange positioning system based on deep learning specifically includes: Through the following formula, Convert the LiDAR point cloud data into a two-dimensional image, where is the position of the point cloud on the image, is the location of the point cloud in the real world, is the transfer matrix from the lidar point cloud position to the image point cloud position, is the offset between the lidar point cloud position and the image point cloud position; By setting the following parameters, one pixel is set to be equal to 0.05m. For the image resolution, the point cloud range is set within a square range of 10m × 10m, with the center of the mobile robot as the base point, and the real scale is converted into image pixels; Through the transformation matrix, the map coordinates (100, 200) are used as the center of the mobile robot to draw the point cloud information on the two-dimensional image; Image edge detection is used to determine the region of interest within the image range of x ≥ 100 and 90 ≤ y ≤ 110, and noise is filtered on the content of the region of interest to obtain a noise-free image; The noise-free image is fed into a deep neural network to determine the corridor area. The improved Inception V3 deep neural network architecture is used as the corridor recognition network. In the corridor classification task, the fine-tuning method is used based on the ImageNet dataset to deploy the deep learning method to the AGV equipment. When the trained deep neural network detects a long corridor area, the AGV switches from the SLAM positioning system to the IMU-based odometry positioning system.

7. A fully autonomous navigation device for an indoor sparse feature mobile robot, characterized in that: include: Trajectory tracking module for predicting turning distance and performing steering actions using the angle a between two vectors created from two points from the current robot center to the next forward-looking point and the current rotation angle b; 2D laser SLAM module, used in unstructured environment scenarios based on long corridors, uses 2D laser radar data and SLAM algorithm to match point clouds with map features. If the positioning is successful, the point cloud will be superimposed on the black edge of the SLAM map, and it can be determined whether the AMR is positioned incorrectly. The system positioning module is used to convert point cloud data into a two-dimensional image through a preset algorithm, draw point cloud information on the two-dimensional image, use the obtained two-dimensional point cloud image as input for model training, and deploy the deep learning method to the AGV equipment, wherein when the trained deep neural network detects a long corridor area, the AGV switches the SLAM positioning system to the IMU-based odometer positioning system.

8. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program implements the method according to any one of claims 1 to 8 when executed by the processor.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images