Railway vehicle carriage inspection method, inspection robot and railway vehicle

Abstract

The invention discloses a railway vehicle carriage inspection method, an inspection robot and a railway vehicle. The method comprises the steps that a three-dimensional data model or a two-dimensional CAD model of a train is acquired, and a navigation system generates a two-dimensional inspection map in a train carriage according to the three-dimensional data model or the two-dimensional CAD model; an inspection path, an inspection position point and an inspection compartment number in a train compartment and a rough position of the inspection compartment number in the inspection map are set in the inspection map; accurate positioning of the inspection robot in the inspection map and the advancing direction of the inspection robot are determined; and the inspection robot performs inspection in each compartment passenger room of the train according to the set inspection task, and obstacle avoidance is performed in the inspection process. According to the invention, through accurate positioning and obstacle avoidance of fusion of the map generated by the train data model and multi-sensor information, high-precision inspection walking of the inspection robot in a narrow channel of a train compartment and positioning of the inspection robot in the same scene of each compartment are realized.

Description

Technical Field

[0001] The present invention relates to the field of vehicle compartment inspection, and in particular to a rail vehicle compartment inspection method, an inspection robot, and a rail vehicle. Background Art

[0002] The statements herein merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In order to ensure the safety of rail transit vehicles and the comfort of passengers, it is necessary to conduct comprehensive inspections on the interior of rail transit vehicle compartments on a regular basis.

[0004] In patent CN202320707103.6, an intelligent inspection robot in a motor vehicle compartment is disclosed; it includes a mobile chassis and a mounting frame mounted on the mobile chassis, the mobile chassis and the mounting frame form a structural setting that is conducive to achieving anti-overturning during travel, the mobile chassis is a mobile chassis that is conducive to climbing and crossing obstacles, and the mobile chassis and the mounting frame are installed with a main control module, a visual detection module, an autonomous obstacle avoidance walking module, a communication module, a warning module, and a power supply module that constitute a robot system for automatic obstacle avoidance walking and intelligent inspection and detection of the mobile chassis. The visual detection module is installed at the upper end of the mounting frame, and the induction working end of the autonomous obstacle avoidance walking module is set against the lower part of the mounting frame. However, the patent does not disclose a specific inspection method, but is mainly aimed at improving the robot itself.

[0005] Patent CN 202210244398.8 also discloses an intelligent rail vehicle inspection robot, including a system platform and an inspection robot. The system platform interacts with the inspection robot through a wireless network, and a mobile base is arranged at the bottom of the inspection robot; the inspection robot is provided with an image acquisition system, a safety protection system and an automatic positioning system; the image acquisition system is used to collect inspection images of rail vehicles, and the image acquisition system includes a robotic arm, and a two-dimensional image acquisition component and / or a three-dimensional image acquisition component mounted on the robotic arm; the robotic arm adopts a six-degree-of-freedom robotic arm; the two-dimensional image acquisition component includes an LED light and a binocular camera; the three-dimensional image acquisition component includes a projector and a binocular camera; the safety protection system includes an ultrasonic radar arranged around the mobile base of the inspection robot, and the ultrasonic radar is used to scan surrounding obstacles; the automatic positioning system adopts a SLAM navigation system. However, the robot uses autonomous navigation technology, high-definition photography technology, multi-eye 3D detection technology, and wireless data transmission technology to automatically detect the bogies and underbody of the urban EMU trains on the maintenance trench in the parking depot; it uses deep learning algorithms to identify and locate key components, such as front-view bolts, side-view bolts, cable joints, etc., to achieve a detection effect based on the status of the components; and through algorithms, it reduces false alarms caused by water stains, stains, and changes in lighting. However, it does not involve the robot's path planning and map generation.

[0006] Some methods for patrol robot inspection are disclosed in other existing technologies, but the commonly used methods for patrol robot map generation are as follows: (1) LiDAR scanning: This is one of the most commonly used methods. LiDAR measures the distance information of the surrounding environment by emitting lasers and receiving reflected signals, and then constructs a three-dimensional map. This method has high accuracy and is suitable for various environments. (2) Visual sensor: The visual sensor generates a map by capturing images of the surrounding environment and processing them using computer vision algorithms. This method works better in environments with sufficient light and rich textures. (3) Fusion of multiple sensors: In order to improve the accuracy and robustness of the map, some advanced robot inspection systems will fuse data from multiple sensors such as LiDAR, visual sensors, and inertial measurement units (IMUs) to generate a more accurate map. However, the problems with the mapping method of patrol robots are as follows:

[0007] (1) Cost issue: High-precision sensors such as lidar are expensive, which increases the overall cost of the robot.

[0008] (2) Poor environmental adaptability: Map generation and updating may face challenges in complex or dynamically changing environments. For example, the quality of map generation may be affected in environments with poor lighting, lack of texture, or a large number of dynamic obstacles.

[0009] (3) Large amount of data processing: Fusion of data from multiple sensors requires powerful computing power and efficient algorithm support, which places high demands on the robot’s processor and data storage capabilities. Summary of the invention

[0010] In view of the deficiencies in the prior art, an object of the present invention is to provide a rail vehicle carriage mobile inspection robot system and an inspection method thereof.

[0011] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0012] In a first aspect, an embodiment of the present invention provides a rail vehicle carriage inspection method, which uses an inspection robot for inspection, and the specific inspection steps are as follows:

[0013] Step 1: The navigation system of the inspection robot obtains a 3D data model or a 2D CAD model of the train; generates a 2D inspection map of the interior of the train compartment based on the imported 3D data model or the 2D CAD model; by generating a map from the model, the robot can plan a path in advance in a complex environment to avoid obstacles and dangerous areas;

[0014] Step 2: Set the inspection path, inspection location, and inspection carriage number in the train carriage and its rough location in the inspection map;

[0015] Step 3: Obtain relevant information collected by sensors on the inspection robot and fuse the relevant information to determine the precise positioning of the inspection robot in the inspection map and the direction of the inspection robot;

[0016] Step 4: The inspection robot inspects each compartment of the train according to the set inspection tasks and avoids obstacles during the inspection process; when the inspection robot moves to the inspection task point of the train compartment, it stops and collects relevant information.

[0017] As a further technical solution, in step 1, the steps of generating a two-dimensional inspection map of the interior of the train compartment according to the three-dimensional data model are as follows:

[0018] Convert the format of the 3D data model;

[0019] Projecting the converted three-dimensional data model onto a two-dimensional plane;

[0020] rasterizing the projected two-dimensional image information;

[0021] Perform data processing on raster information;

[0022] Construct a two-dimensional inspection map.

[0023] As a further technical solution, data processing of grid information includes height information extraction, state determination, and denoising and smoothing.

[0024] As a further technical solution, a method for constructing a two-dimensional inspection map is as follows: Create a two-dimensional array or matrix corresponding to the grid grid to store the grid map. Fill the corresponding position in the map according to the state of each grid; usually, the occupied state is represented by a specific color or value, and the idle state is represented by another color or value; then, the generated map is optimized and post-processed;

[0025] As a further technical solution, the generated map is optimized and post-processed: the map is denoised to eliminate error states caused by data noise or projection errors; the map is smoothed to improve the visual effect and continuity of the map; the map boundaries are processed as needed to ensure the integrity and accuracy of the map.

[0026] As a further technical solution, a method for constructing a two-dimensional inspection map is as follows:

[0027] Use map rendering to convert raster data into a visual two-dimensional map; add annotations and legends to the map as needed; and manage the map layers;

[0028] Convert the generated two-dimensional map into a map format for inspection robot navigation;

[0029] Use the tool to connect to the inspection robot and upload the converted map file to the robot's backend database system;

[0030] Configure map parameters in the robot navigation system to ensure that the robot can accurately identify and use the map;

[0031] Use real data or simulated data to verify the accuracy of the generated 2D map to ensure that it meets the needs of inspection robot navigation applications.

[0032] As a further technical solution, in step 1, the steps of generating a two-dimensional inspection map of the interior of the train compartment according to the two-dimensional CAD model are as follows:

[0033] Convert 2D CAD models to DXF format files;

[0034] Convert the format file into raster data used by the passenger room inspection robot for navigation; during the conversion process, it is necessary to extract the geometric information in the 2D CAD model, which will be used to create corresponding black fills (representing obstacles / boundary areas) and white fills (representing blank areas / usable areas) in the raster map.

[0035] Generate a grid map, use ROS to generate a grid map, write a ROS node that publishes a grid map message of the interior of the passenger compartment and sets the resolution, width, and height of the map. Then, based on the geometric information in the CAD drawing, assign a value to each grid cell, indicating the probability that the cell is occupied by an obstacle.

[0036] Visualize 2D inspection maps in Rviz.

[0037] As a further technical solution, in step 3, the point cloud data scanned by the laser radar on the inspection robot, the information collected by the ultrasonic obstacle avoidance sensor, and the image information collected by the visible light camera are fused; the fusion method is as follows:

[0038] Preprocess the point cloud data scanned by the LiDAR, the information collected by the ultrasonic obstacle avoidance sensor, and the image information collected by the visible light camera;

[0039] Then, the pre-processed sensor data is time synchronized; the timestamps of each sensor are calibrated and the time correspondence between them is found. Time synchronization can ensure that the subsequent data fusion and processing are based on information at the same time point, thereby improving the accuracy and reliability of the system.

[0040] Then, the time-synchronized sensor data is spatially synchronized, and the data collected by different sensors are aligned according to the spatial position to ensure that they describe the environmental state within the same spatial range. (Coordinate transformation and registration are performed on the data of each sensor to convert them to a unified coordinate system. Through spatial synchronization, it can be ensured that the subsequent data fusion and processing are based on information within the same spatial range, thereby further improving the accuracy and reliability of the system.)

[0041] Finally, the sensor data are fused after being synchronized in time and space.

[0042] As a further technical solution, in step 6, the inspection robot adjusts the mechanical arm to the optimal arm shape according to the state of the mechanical arm set in advance, and the detection sensor collects relevant information at the optimal height and angle;

[0043] As a further technical solution, after the relevant information is collected in step 7, the collected information is transmitted to the background monitoring system through the 5G / WiFi wireless module for processing and analysis, and the analysis results are fed back to the robot local control platform;

[0044] As a further technical solution, during inspection, it is necessary to clearly define the task points of each inspection task; during the inspection process, the position of the inspection robot is located in real time through information such as odometer, inertial sensor, lidar, etc., and the information fusion of lidar and ultrasonic obstacle avoidance sensor is used to avoid obstacles in the passenger compartment environment (seats, etc.) of the train compartment.

[0045] As a further technical solution, the odometer is used to provide the robot's linear velocity and angular velocity information, and preliminarily estimate the robot's position and posture; the inertial sensor is used to provide the robot's instantaneous acceleration and angular velocity information, which helps to correct the position drift problem caused by the odometer's integration error;

[0046] By adding the odometer information and lidar point cloud information output by the inertial sensor fusion output, a 3D point cloud map of the carriage is established, which is used as the positioning priori information of the passenger compartment inspection robot during the automatic inspection process;

[0047] A relocalization strategy based on a priori 3D point cloud maps is adopted. The input is lidar information, inertial sensor information, and odometer information after inertial sensor fusion. The output is the relative position information of the inspection robot in the 3D point cloud map of the train compartment.

[0048] Ultrasonic obstacle avoidance sensors are used to provide close-range obstacle detection, while lidar provides long-range obstacle detection, thus enabling obstacle detection and early warning during the operation of the inspection robot.

[0049] In a second aspect, an embodiment of the present invention further provides a rail vehicle, the carriages of which are inspected using the inspection method described above.

[0050] The beneficial effects of the above embodiments of the present invention are as follows:

[0051] The method proposed in the present invention for the inspection robot to generate a map based on a two-dimensional or three-dimensional model and realize navigation and positioning in combination with sensors combines the advantages of model prediction and real-time perception and has many advantages.

[0052] (1) The present invention obtains a three-dimensional data model or a two-dimensional CAD model of the train; generates a two-dimensional inspection map of the interior of the train compartment according to the imported three-dimensional data model or two-dimensional CAD model; eliminates the high cost of collecting data through a high-precision laser radar sensor during the mapping stage and then building a map, thereby reducing costs. At the same time, since the inspection map of the present application is a two-dimensional inspection map generated entirely based on an existing train model, there will be no problems of poor environmental adaptability caused by dynamic changes and insufficient light, and there is no need to fuse data from multiple sensors for mapping, thereby reducing problems such as large data processing volume;

[0053] (2) High-precision positioning and navigation

[0054] Fusion of multiple information sources: This method combines the map information generated by the model and the data perceived by the sensor in real time to provide a more comprehensive and accurate description of the environment. This helps the robot achieve high-precision positioning and navigation.

[0055] Adapt to complex environments: By generating maps through models, robots can plan paths in complex environments in advance and avoid obstacles and dangerous areas. At the same time, the data perceived by sensors in real time can help robots respond to emergencies and environmental changes, and improve the flexibility and safety of navigation.

[0056] (3) Enhanced robustness and reliability

[0057] Redundant design: There is a redundant relationship between the model-generated map and the sensor perception data. Even if one information source fails or has errors, the other information source can still provide effective navigation information, enhancing the robustness and reliability of the system.

[0058] Continuous learning and optimization: Model-based map generation can continuously optimize the accuracy and completeness of the map through continuous learning. At the same time, sensor data can also be used to correct maps and navigation algorithms in real time, improving the overall performance of the system.

[0059] (4) Improve inspection efficiency and accuracy

[0060] Autonomous navigation and obstacle avoidance: Through the fusion of model-generated maps and sensor perception data, the robot can achieve autonomous navigation and obstacle avoidance and complete inspection tasks without human intervention. This greatly improves the efficiency and accuracy of inspections.

[0061] Real-time monitoring and early warning: Sensors can monitor equipment status and environmental parameters in real time. Once an abnormal situation or potential risk is detected, the robot can immediately issue an early warning and take appropriate countermeasures. This helps to detect and deal with problems in a timely manner, ensuring the safety and effectiveness of inspections.

[0062] (5) Reduce costs and manpower requirements

[0063] Reduced human intervention: Based on the fusion of model-generated maps and sensor perception data, the robot can complete inspection tasks autonomously, reducing the need for human intervention. This helps reduce labor costs and improve work efficiency.

[0064] Improve equipment utilization: Through autonomous navigation and obstacle avoidance, the robot can move flexibly in complex environments and cover a wider area. This helps to improve equipment utilization and inspection coverage.

[0065] (6) Expand application areas and scenarios

[0066] Adapt to various environments: This method is applicable to various environments and scenarios, such as factories, warehouses, substations, tunnels, etc. By adjusting the model and algorithm parameters, the robot can adapt to different inspection needs and environmental characteristics. Support multiple tasks: Based on the fusion of model-generated maps and sensor perception data, the robot can not only complete inspection tasks, but also support monitoring, rescue, detection and other tasks. This helps to expand the application areas and scenario range of robots. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0068] Figure 1 It is a structural schematic diagram of the inspection robot proposed by the present invention;

[0069] Figure 2 It is a structural schematic diagram of the inspection robot proposed by the present invention;

[0070] Figure 3 It is a structural schematic diagram of the inspection robot proposed by the present invention;

[0071] Figure 4 It is a structural schematic diagram of the inspection robot proposed by the present invention;

[0072] Figure 5 It is a schematic diagram of the working process of the inspection robot proposed by the present invention;

[0073] In the figure: the distances or sizes between parts are exaggerated to show the positions of various parts, and the schematic diagram is for illustration only.

[0074] 1. Car body, 2. Modular hub motor module, 3. Universal wheel, 4. LiDAR, 5. Lightweight collaborative robotic arm, 6. Ultrasonic obstacle avoidance sensor, 7. Dual-view camera, 8. Automatic charging brush plate. DETAILED DESCRIPTION

[0075] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0076] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless otherwise explicitly stated in the present invention, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "include" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof;

[0077] As introduced in the background technology, there are deficiencies in the prior art. In order to solve the above technical problems, the present invention proposes a rail vehicle carriage mobile inspection robot system and an inspection method thereof.

[0078] In a typical embodiment of the present invention, Figure 1 , Figure 2 , Figure 3 , Figure 4 As shown, the inspection robot provided in this embodiment is composed of a mobile chassis, a light collaborative robot arm 5 and a detection sensor. The mobile chassis is composed of a vehicle body 1, a modular hub motor module 2, a universal wheel 2, a controller, a laser radar 4, an ultrasonic obstacle avoidance sensor 6, a 5G / Wifi wireless communication module, etc., which is responsible for walking in the passenger compartment of the train and navigating and avoiding obstacles; the light collaborative robot arm 5 is installed on the fixed mounting plate of the mobile chassis, and is responsible for adjusting the posture of the terminal detection sensor (dual-view camera 7) to ensure the best height and angle for detection and photography.

[0079] The detection sensor (dual-view camera 7) is installed at the end of the mechanical arm, which integrates an infrared thermal imaging camera and a high-definition visible light camera. The infrared thermal imaging camera is responsible for the temperature detection of the switchboard in each carriage of the train to determine whether there is a fault; the high-definition visible light camera is responsible for the detection of electronic seat numbers, light strips, LCD screens and LED electronic screens. The inspection robot is equipped with ultrasonic obstacle avoidance sensors 6 on both sides of the body, which are mainly placed on the mobile robot and the seats in the passenger compartment of the train compartment during the movement. The controller is the brain of the entire inspection robot, which is mainly responsible for the motion control of the mobile chassis and the mechanical arm, the collection and processing of sensor information such as the navigation and positioning system, and information communication with the background monitoring system of the train inspection.

[0080] The inspection robot is also equipped with an automatic charging brush plate 8, which can autonomously walk to the charging pile and autonomously charge. The inspection robot navigation control system of the present invention is composed of a laser radar sensor, an ultrasonic obstacle avoidance sensor, an inertial sensor, an odometer, and a visual sensor. The sensor fusion of the navigation system can accurately determine the position of the inspection robot in the inspection map, and can ensure that the inspection robot moves in narrow passages and avoids collisions with the external environment, especially seats.

[0081] The inspection robot disclosed in this embodiment uses a 5G+Wifi6 dual-frequency module as the core functional module of communication. The wireless module is connected to the controller, visual sensor, etc. through a wired network via a connected switch, and communicates with the 5G or WiFi communication base stations in each workshop of the factory through 5G+Wifi6 wireless communication. The base station communicates with the background monitoring system through optical fiber communication.

[0082] The initial positioning of the mobile inspection robot is achieved through the point cloud data information scanned by its lidar, the information collected by the ultrasonic obstacle avoidance sensor, and the image information collected by the visible light camera on the lightweight collaborative robot in the four directions of the car, and the manually set initial position is calibrated by fusing the relevant information.

[0083] During the inspection process, the visual image information is used to confirm the car it is in, its rough position in the car and the direction of travel. The odometer, inertial sensor, lidar and other information are used to locate the inspection robot in real time. The information fusion of lidar and ultrasonic obstacle avoidance sensor is used to avoid obstacles in the passenger compartment environment (seats, etc.) of the train compartment, and to avoid obstacles such as people walking in the compartment. When the inspection robot is less than the safe distance between the moving object and the inspection robot (set as needed), the inspection robot will stop moving until the distance between the moving object and the inspection robot is greater than the safe distance, at which time the inspection robot will stop moving and continue the inspection task.

[0084] Furthermore, based on the above inspection robot system, the inspection and navigation workflow proposed in this embodiment is as follows:

[0085] Step 1: Import the 3D data model or 2D CAD model of the train design into the inspection robot navigation software;

[0086] Step 2 generates a two-dimensional map of the interior of the train compartment based on the imported three-dimensional data model or two-dimensional CAD model;

[0087] Furthermore, in step 2, the method of generating a two-dimensional map of the interior of a train compartment from the three-dimensional data model of the EMU train is as follows:

[0088] Step 2-1-1 Create or obtain the 3D data model of the EMU train

[0089] When using 3D modeling software to create or obtain a 3D data model of an EMU train, it is necessary to ensure the accuracy and completeness of the 3D model information of the entire EMU train, including various information such as seats in the carriage, display screen locations, train access doors, etc.

[0090] Step 2-1-2 Format conversion of 3D data model. Convert the 3D data model of the EMU train into a common 3D format (such as OBJ, STL, FBX, etc.) through modeling software such as Inventor, 3ds Max, Solidworks, etc.

[0091] Step 2-1-3 Project the 3D data model onto a 2D plane. Use 3D modeling software or professional map projection tools to orthogonally project the 3D data model onto a 2D plane. In this process, you need to set parameters such as the projection plane, projection center, and projection direction.

[0092] Step 2-1-4 rasterizes the projected two-dimensional image information. Use rasterization tools or algorithms to divide the projected two-dimensional image into a series of regular grid cells. Each grid cell can contain attribute information such as height, color, and material. During the rasterization process, the size of the grid is determined based on the accuracy requirements and computing resources for creating the map. The smaller the grid, the higher the accuracy of the map, but the greater the amount of calculation.

[0093] Step 2-1-5: Grid data information processing: Data processing is performed on the grid information obtained in the previous step, including height information extraction, state determination, denoising and smoothing.

[0094] ① High information extraction

[0095] For each grid cell, the height information in the corresponding three-dimensional data model can be extracted, which can be achieved by methods such as interpolation calculation.

[0096] ② Status determination

[0097] Based on the height information and other conditions (such as threshold settings), the state of each grid cell (such as occupied, idle, etc.) is determined.

[0098] ③ Denoising and smoothing:

[0099] It is necessary to denoise the raster data to eliminate the influence of noise points. At the same time, the raster data should be smoothed to improve the visual effect and continuity of the map.

[0100] Step 2-1-6 Map construction method

[0101] First method:

[0102] Create a two-dimensional array or matrix corresponding to the grid grid to store the grid map. Fill the corresponding position in the map according to the state of each grid. Usually, the occupied state is represented by a specific color or value, and the idle state is represented by another color or value. Then, optimize and post-process the generated map. The main process is as follows:

[0103] Denoising: Denoising is performed on the map to remove erroneous states caused by data noise or projection errors.

[0104] Smooth: Smooths the map to improve the visual quality and continuity of the map.

[0105] Boundary processing: Process the map's boundaries as needed to ensure map integrity and accuracy.

[0106] Second method:

[0107] Use map rendering tools such as ArcGIS or Mapbox Studio or Mapnik to convert raster data into a visual two-dimensional map. This includes assigning different colors, textures, etc. to different raster states.

[0108] As needed, add annotations, legends and other information to the map so that users can better understand the map content. At the same time, manage the map in layers to facilitate subsequent editing and updating.

[0109] Format conversion of generated maps. Convert the generated two-dimensional maps in DWG, DXF, Shapefile and other formats into map formats for inspection robot navigation, such as Pcap files, Las files, Tilemap files, etc.

[0110] Map upload: Use the tool to connect to the inspection robot and upload the converted map file to the robot's backend database system.

[0111] Map configuration: Configure map parameters in the robot navigation system, such as zoom level, positioning accuracy, etc., to ensure that the robot can accurately recognize and use the map.

[0112] Map validation and testing

[0113] Use real data or simulated data to verify the accuracy of the generated 2D map to ensure that it meets the needs of inspection robot navigation applications.

[0114] Furthermore, in step 2, the process of converting the CAD image of the interior of the high-speed rail passenger compartment into a grid map used by the inspection robot for navigation can be divided into several steps, involving parsing the compartment CAD file, extracting geometric information, and converting it into occupancy probability values ​​in the grid map. The following is a specific implementation method:

[0115] Step 2-2-1. First convert the CAD file to DXF format: Save the CAD file as DXF format, because DXF is an open file format and is easier to be read and processed by other software.

[0116] Step 2-2-2. Use conversion tools: You can use conversion tools of GIS software such as ArcGIS to convert DXF files into raster data used for navigation of the passenger room inspection robot.

[0117] Step 2-2-3. Extract geometric information: During the conversion process, it is necessary to extract geometric information from the CAD image, such as lines and shapes. This geometric information will be used to create corresponding black fills (representing obstacles / boundary areas) and white fills (representing blank areas / exerciseable areas) in the raster map.

[0118] Step 2-2-4. Generate a grid map: Specifically, you can use the "nav_msgs / OccupancyGrid" message type in ROS to generate a grid map. Write a ROS node that publishes a grid map message of the interior of the passenger compartment and sets the resolution, width, and height of the map. Then, based on the geometric information in the CAD drawing, assign a value to each grid cell, indicating the probability that the cell is occupied by an obstacle.

[0119] Step 2-2-5 Visualize in Rviz 3D visualization platform: Once the raster map is generated, you can add a Map component in Rviz, a visualization tool in ROS, to visualize it. Set the correct "Fixed Frame" and "Topic" so that Rviz can receive and display the raster map data in the car.

[0120] Step 3: Set the inspection path and inspection location points in the train compartment in the inspection map;

[0121] The 2D grid map of the carriage generated by the CAD file in the previous step is used for the navigation of the inspection robot. When the inspection robot can locate normally using the map, the remote-controlled robot moves in the carriage and records the robot's point information (the robot's relative position information in the grid map, including position and heading (x, y, theta)) at the corresponding inspection point for automatic inspection by the robot.

[0122] Step 4 Before the inspection, set the carriage number where the inspection robot starts to inspect and its rough position in the map;

[0123] Step 5: The inspection robot can accurately locate the mobile inspection robot in the map and determine the direction of the inspection robot based on the multi-sensor information fusion such as the laser radar sensor, ultrasonic obstacle avoidance sensor, inertial sensor, odometer and visual sensor;

[0124] Furthermore, the point cloud data scanned by the lidar, the information collected by the ultrasonic obstacle avoidance sensor, and the image information collected by the visible light camera on the lightweight collaborative robot are fused, which usually involves the following steps:

[0125] Step 5-1 Data preprocessing: First, the point cloud data of the lidar, the data of the ultrasonic sensor, and the image information collected by the visible light camera need to be preprocessed; the processing process includes filtering, denoising, calibration and other steps to ensure the accuracy and availability of the data.

[0126] Step 5-2: Time synchronization of the preprocessed data in step 5-1: Since the data acquisition frequency of different sensors may be different, the data needs to be time synchronized to ensure that all sensor data are consistent in time. Use message filters (message_filters) in ROS (Robot Operating System) to achieve soft synchronization of multi-sensor data.

[0127] Step 5-3 performs spatial alignment on the time-synchronized data in step 5-2: data from different sensors need to be mapped to the same coordinate system. This usually involves the calibration of the sensor's external parameters, that is, determining the position and orientation of the sensor in the robot's coordinate system. Ultrasonic obstacle avoidance sensors mainly play a role in filling in blind spots in the passenger compartment inspection robot. The core sensors used for the perception of the environment information in the compartment are cameras and lidars. It is necessary to calibrate their internal and external parameters and align their coordinate systems to the coordinate system of the inspection robot itself.

[0128] Step 5-4 performs data fusion: Different strategies can be used for data fusion. The point cloud data can be projected onto the image plane to obtain the RGB values ​​of the corresponding pixels, and the point cloud of the lidar can be colored. This allows the passenger room inspection robot to more accurately confirm the obstacle information around the robot.

[0129] Step 5-5 Target detection and classification: The fused data can be used for target detection and classification. Use Yolo (machine vision tool) to identify objects in the image, and combine the information output by Yolo with the data from the lidar and ultrasonic sensors to improve the accuracy of detection.

[0130] Step 6: The inspection robot inspects each compartment of the train according to the set inspection tasks. During the inspection, the task points of each inspection task need to be clearly defined. During the inspection, the inspection robot is located in real time through information such as odometer, inertial sensor, and lidar, and the internal environment of the train compartment (seats, etc.) is avoided through information fusion of lidar and ultrasonic obstacle avoidance sensor. Furthermore, the obstacle avoidance process of this step is as follows:

[0131] Step 6-1 The odometer of the passenger compartment inspection robot is used to provide the robot's linear velocity and angular velocity information, and preliminarily estimate the robot's position and posture. Since the odometer adopts a velocity integration strategy, there is a certain cumulative error. At this time, an inertial sensor (IMU) is added to provide the robot's instantaneous acceleration and angular velocity information, which helps to correct the odometer's position drift caused by the integration error.

[0132] Step 6-2 uses the odometer information and lidar point cloud information output by adding IMU fusion to establish a 3D point cloud map inside the car, which is used as a priori positioning information for the passenger compartment inspection robot during the automatic inspection process.

[0133] Step 6-3 Inspection robot positioning: adopt a repositioning strategy based on a priori 3D point cloud map, with the input being lidar information, IMU information, and odometer information after IMU fusion, and the output being the relative position information of the inspection robot in the 3D point cloud map of the train compartment.

[0134] In step 6-4, the laser radar detects obstacles at close range and long range, and the ultrasonic obstacle avoidance sensor detects obstacles at close range. Since the ultrasonic obstacle avoidance sensor has a greater advantage in close range than the laser radar, it can provide close-range obstacle detection to supplement the blind spot of the laser radar at close range. This enables the inspection robot to detect and warn of obstacles during operation, and then transmits the detection information to the control module of the inspection robot. The inspection robot control module performs path planning and adjustment based on the position information obtained in step 6-3 and the obstacle information obtained by the laser radar and ultrasonic obstacle avoidance sensor, or controls the robot to stop and wait for the obstacle to be removed.

[0135] Step 7 The inspection robot stops at the train compartment inspection task point. According to the pre-set state of the robotic arm, the robotic arm is adjusted to the optimal arm shape, and the detection sensor collects relevant information at the optimal height and angle.

[0136] Step 8 After each test, the test results are transmitted to the background monitoring system through the 5G / WiFi wireless module for processing and analysis, and the analysis results are fed back to the robot local control platform;

[0137] Step 9: The staff confirms the relevant faults based on the fault information fed back on the local control platform and performs further maintenance.

[0138] In this embodiment, the initial positioning of the mobile inspection robot is performed by using the point cloud data information scanned by its laser radar, the information collected by the ultrasonic obstacle avoidance sensor, and the image information collected by the visible light camera on the light collaborative robot in the four directions of the front, back, left, and right in the carriage, and the manually set initial position is calibrated by fusing relevant information. During the inspection process, the visual image information is used to confirm the carriage in which it is located, as well as the rough position and direction of travel in the carriage. The position of the inspection robot is located in real time through information such as the odometer, inertial sensor, and laser radar. The information fusion of the laser radar and the ultrasonic obstacle avoidance sensor is used to avoid obstacles in the passenger compartment environment (seats, etc.) of the train compartment, and to avoid obstacles for moving objects such as people walking in the compartment. When the inspection robot is less than the safe distance from the moving object (set as needed), the inspection robot will stop moving until the distance between the moving object and the inspection robot is greater than the safe distance, and the inspection robot will release the stopped motion state and continue the inspection task. The inspection robot uses a 5G+Wifi6 dual-frequency module as the core functional module for long-distance communication. The wireless module communicates with the controller, visual sensor, lidar, etc. through a switch inside the mobile robot via a wired network, and communicates with the 5G or WiFi communication base stations in each workshop of the factory through 5G+Wifi6 wireless communication. The base station communicates with the background monitoring system of train inspection through a fiber optic network.

[0139] This embodiment also provides a patrol robot, which is a special environment application robot, and its particularity of detection scene is particularly reflected in the complexity of the carriage environment. The patrol robot of the present invention is similar to the traditional patrol robot in structure, and the biggest difference lies in the navigation and positioning method. The patrol robot adopts a standard design model (three-dimensional design model or two-dimensional design model) in the navigation and positioning system, and generates a two-dimensional map that can be directly used by the patrol robot by converting and processing the model. This method can reduce the work of manual mapping required by traditional patrol methods, and can directly use the generated two-dimensional map to directly navigate the patrol robot, and can set the task points and the turning and U-turn positions of the patrol robot on the generated two-dimensional map.

[0140] The above inspection method can also solve a series of problems faced by inspection robots in rail vehicle passenger compartment inspection, such as precise movement and obstacle avoidance in narrow channels, and loss of navigation position in long channels with similar scenes in each carriage. The inspection robot generates a map based on a two-dimensional or three-dimensional model, combines sensors to achieve precise navigation positioning and obstacle avoidance, and then accurately controls the movement of the inspection robot to effectively solve the problem of precise movement in narrow channels. In a narrow channel, the inspection robot can adjust its speed and steering angle according to the width and curvature of the channel to avoid collision and maintain a stable motion trajectory. Obstacles in the surrounding environment are detected in real time through sensors, and obstacle avoidance decisions are made in combination with map information. A feedback mechanism is established to monitor the robot's motion state and environmental changes in real time. When an abnormal situation is detected or the predetermined trajectory is deviated, the robot can adjust its motion strategy or issue an alarm signal in time.

[0141] When navigating in long passages, inspection robots may encounter the problem of lost positioning, which is mainly caused by factors such as the single environment of the long passage, lack of feature points, or signal interference. To solve this problem, inspection robots can generate maps based on two-dimensional or three-dimensional models and combine sensors to achieve accurate navigation positioning and obstacle avoidance methods.

[0142] During the inspection process, the robot should continuously collect environmental data and compare and integrate it with the initial model-based generated map to update the map information in real time.

[0143] When updating the map, key feature points in the channel are extracted, such as corners, doorways, obstacles, etc. These feature points help the robot maintain its positioning in long channels.

[0144] Furthermore, the present embodiment also provides a rail vehicle, the rail vehicle comprising the rail vehicle carriage mobile inspection robot system and the inspection method thereof as described above. Since the rail vehicle is provided with the rail vehicle carriage mobile inspection robot system and the inspection method thereof as described above, the rail vehicle also has all the advantages as described above. In some embodiments, the rail vehicle provided by the present invention can be any appropriate type of vehicle, such as a conventional train, a motor vehicle, a subway vehicle, a city rail vehicle, etc. The present invention is not limited to one or some specific rail vehicle types.

[0145] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A rail vehicle carriage inspection method, using an inspection robot for inspection, characterized in that: The inspection steps are as follows: Step 1: Obtain a three-dimensional data model or a two-dimensional CAD model of the train, and the navigation system generates a two-dimensional inspection map of the interior of the train compartment according to the three-dimensional data model or the two-dimensional CAD model; Step 2: Set the inspection path, inspection location, and inspection carriage number in the train carriage and its rough location in the inspection map; Step 3: Obtain relevant information collected by sensors on the inspection robot and fuse the relevant information to determine the precise positioning of the inspection robot in the inspection map and the direction of the inspection robot; Step 4: The inspection robot inspects each compartment of the train according to the set inspection tasks and avoids obstacles during the inspection process; when the inspection robot moves to the inspection task point of the train compartment, it collects relevant information.

2. The rail vehicle compartment inspection method according to claim 1, characterized in that: In step 1, the steps of generating a two-dimensional inspection map of the interior of the train compartment according to the three-dimensional data model are as follows: Convert the format of the 3D data model; Projecting the converted three-dimensional data model onto a two-dimensional plane; rasterizing the projected two-dimensional image information; Perform data processing on raster information; Construct a two-dimensional inspection map.

3. The rail vehicle compartment inspection method according to claim 2, characterized in that: Method for constructing a two-dimensional inspection map: create a two-dimensional array or matrix corresponding to the grid grid to store the grid map; fill the corresponding position in the map according to the state of each grid; usually, the occupied state is represented by a specific color or value, and the idle state is represented by another color or value; then, optimize and post-process the generated map.

4. The rail vehicle compartment inspection method according to claim 2, characterized in that: The process of optimizing and post-processing the generated map: denoising the map to eliminate erroneous states caused by data noise or projection errors; smoothing the map to improve the visual effect and continuity of the map; Manipulate the map's boundaries as necessary to ensure the map's integrity and accuracy.

5. The rail vehicle compartment inspection method according to claim 2, characterized in that: The method of constructing a two-dimensional inspection map is as follows: Use map rendering to convert raster data into a visual two-dimensional map; add annotations and legends to the map as needed; and manage the map layers; Convert the generated two-dimensional map into a map format for inspection robot navigation; Use the tool to connect to the inspection robot and upload the converted map file to the robot's backend database system; Configure map parameters in the robot navigation system to ensure that the robot can accurately identify and use the map; Use real data or simulated data to verify the accuracy of the generated 2D map to ensure that it meets the needs of inspection robot navigation applications.

6. The rail vehicle compartment inspection method according to claim 1, characterized in that: In step 1, the steps of generating a two-dimensional inspection map of the interior of a train compartment according to the two-dimensional CAD model are as follows: Convert 2D CAD models to DXF format files; Convert the format file into raster data used by the passenger room inspection robot navigation; during the conversion process, extract the geometric information in the 2D CAD model, and the geometric information will be used to create corresponding black fill and white fill in the raster map, where the black fill represents obstacles / boundary areas and the white fill represents blank areas / exerciseable areas; Generate a grid map using ROS, publish the grid map message of the interior of the passenger compartment using the ROS node, and set the resolution, width, and height of the map; assign a value to each grid cell according to the geometric information in the CAD drawing, indicating the probability that the cell is occupied by an obstacle; Visualize 2D inspection maps in the Rviz 3D visualization platform.

7. The rail vehicle compartment inspection method according to claim 1, characterized in that: In step 3, the point cloud data scanned by the laser radar on the inspection robot, the information collected by the ultrasonic obstacle avoidance sensor, and the image information collected by the visible light camera are fused. The fusion method is as follows: First, the point cloud data scanned by the lidar, the information collected by the ultrasonic obstacle avoidance sensor, and the image information collected by the visible light camera are preprocessed; Then, the preprocessed sensor data is time synchronized; Then perform spatial synchronization on the time-synchronized sensor data; Finally, the sensor data are fused after being synchronized in time and space.

8. The rail vehicle compartment inspection method according to claim 1, characterized in that: During the inspection process, the inspection robot is located in real time through the odometer, inertial sensor, and lidar information, and obstacles are avoided in the passenger compartment environment of the train compartment through the fusion of information from the lidar and ultrasonic obstacle avoidance sensor.

9. The rail vehicle compartment inspection method according to claim 8, characterized in that: The obstacle avoidance method is as follows: The odometer is used to provide the robot's linear velocity and angular velocity information, and preliminarily estimate the robot's position and posture; the inertial sensor is used to provide the robot's instantaneous acceleration and angular velocity information, which helps to correct the position drift problem caused by the integration error of the odometer; By adding the odometer information and lidar point cloud information output by the inertial sensor fusion output, a 3D point cloud map of the carriage is established, which is used as the positioning priori information of the passenger compartment inspection robot during the automatic inspection process; A relocalization strategy based on a priori 3D point cloud maps is adopted. The input is lidar information, inertial sensor information, and odometer information after inertial sensor fusion. The output is the relative position information of the inspection robot in the 3D point cloud map of the train compartment. Ultrasonic obstacle avoidance sensors are used to provide close-range obstacle detection, while lidar provides long-range obstacle detection, enabling real-time detection and early warning of obstacles during the operation of the inspection robot.

10. A patrol robot, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the inspection method according to any one of claims 1 to 9 when executing the computer program.

11. A rail vehicle, characterized in that: The carriage is inspected by the inspection method described in any one of claims 1 to 9.

Images