Autonomous getting-off method of robot, robot and system

Abstract

The invention provides a robot autonomous getting-off method, a robot and a system, and relates to the technical field of robot control. The method comprises the steps that in response to a getting-off instruction, a drivable area from a current position to a vehicle door position is determined; controlling a robot body to move towards a vehicle door in the driving area; collecting a first depth image in the direction of a vehicle door, and performing vehicle door edge detection on the first depth image to obtain the position of the lower edge of the vehicle door; and controlling the robot body to get off according to the detected position of the lower edge of the vehicle door. The invention provides a perfect solution for autonomous getting-off of the robot, the adaptability and autonomy of the robot under various terrains are enhanced, and therefore higher operation reliability and safety are achieved.

Description

Technical Field

[0001] The present application relates to the field of robot control technology, and in particular to a method for autonomously getting off a vehicle, a robot and a system. Background Art

[0002] A robot is an intelligent machine that can work semi-autonomously or fully autonomously. A robot can perform tasks such as work or movement through programming and automatic control.

[0003] With the development of intelligent driving technology, robots can perform operations such as inspection and driving on vehicles. However, after the robot completes the operation, it cannot get off the vehicle autonomously, and personnel need to manually lift the robot off the vehicle, which increases labor costs. In view of this, the present application provides a method, robot and system for the robot to get off the vehicle autonomously. Summary of the invention

[0004] In response to the above problems, the present application proposes a method, a robot and a system for autonomously getting off a vehicle, so as to realize autonomously getting off a vehicle by the robot.

[0005] In a first aspect, an embodiment of the present application provides a method for a robot to autonomously get off a vehicle, comprising: In response to the vehicle exit instruction, determining a drivable area from a current position to a vehicle door position; In the drivable area, controlling the robot body to move toward the vehicle door; Collecting a first depth image in the direction of the vehicle door, and performing door edge detection on the first depth image to obtain the position of the lower edge of the vehicle door; According to the detected position of the lower edge of the door, the robot body is controlled to get off the vehicle.

[0006] In a second aspect, an embodiment of the present application provides a robot, comprising: a controller and a robot body connected to the controller; The robot body is equipped with a sensing device for collecting images in the direction of the vehicle door; The controller is used to respond to a get-off instruction and determine a drivable area from a current position to a door position; within the drivable area, control the robot body to move toward the door; collect a first depth image in the direction of the door, and perform door edge detection on the image to obtain the position of the lower edge of the door; and control the robot body to get off the vehicle according to the detected position of the lower edge of the door.

[0007] In a third aspect, an embodiment of the present application provides a robot autonomous getting off a vehicle system, comprising: a robot and a vehicle as described in any embodiment; the robot is located in the vehicle; and the vehicle is used to issue a getting off command to the robot.

[0008] Compared with the prior art, this application has the following technical effects: 1. In this application, the on-board robot performs environmental perception and positioning, determines the drivable area toward the door and automatically moves to the door; then detects the door position and controls the robot body to get off the vehicle, providing a complete solution for the robot to get off the vehicle autonomously.

[0009] 2. This application ensures the safety of the robot's movement by detecting the drivable area and the lower edge of the door, enhances the robot's adaptability and autonomy in various terrains, and thus achieves higher operational reliability and safety.

[0010] 3. The solution provided in this application mainly relies on the sensing device on the robot body to make control decisions, avoiding dependence on external signals and improving the independence and flexibility of the system.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above description and other purposes, features and advantages of the present application more obvious and easy to understand, the preferred embodiments are specifically cited and described in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings: Figure 1 A flowchart of a method for a robot to autonomously get off a vehicle provided in an embodiment of the present application; Figure 2 A schematic diagram of a safety map provided in an embodiment of the present application; Figure 3 A flowchart of another method for a robot to autonomously get off a vehicle provided in an embodiment of the present application; Figure 4 A 2.5D grid image provided in an embodiment of the present application; Figure 5 A schematic diagram of a sector area provided in an embodiment of the present application; Figure 6 A 2D occupancy grid map provided in an embodiment of the present application; Figure 7 A schematic diagram of the structure of the robot provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0014] In the description of this application, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be connected, detachably connected, or integrated; it can be mechanically connected or electrically connected; it can be directly connected or indirectly connected through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0015] The embodiment of the present application provides a method for a robot to autonomously get off a vehicle, which is applicable to a situation where a robot in a vehicle autonomously gets off a vehicle when there is a need to get off the vehicle, without the need for human transport. The method provided in this embodiment is executed by a controller, which may be composed of hardware and / or software. The controller may be integrated in the robot body or electrically connected to the robot body.

[0016] See also Figure 1 , the method provided in this embodiment comprises the following steps: S110: In response to the instruction to get off the vehicle, determine a drivable area from the current position to the door position.

[0017] The vehicle door opens and sends a get-off command to the robot, for example, a "get-off" voice is sent to the robot through the vehicle speaker. The robot collects the "get-off" voice and parses it to obtain the get-off command.

[0018] When the robot receives the command to get off the vehicle, it may not be facing the direction of the vehicle door, so it is necessary to first determine the direction and position of the vehicle door. Optionally, in response to the command to get off the vehicle, a second depth image around the robot body is collected; and the direction and position of the vehicle door are determined based on the depth data in the second depth image.

[0019] In some examples, a depth camera is installed on the robot body, and the lens of the depth camera can be facing upward from the horizontal to detect most of the door features and human features. Since the field of view of the depth camera is limited, the robot body can be controlled to circle in place and continuously shoot to obtain multiple second depth images around the robot body. The depth data in the second depth image represents the distance to the current position. Since the space outside the door is vast, the rate of change of the depth data of each adjacent pixel can be calculated, and the pixel position with the largest rate of change is determined as the door position. Combined with the position and posture of the depth camera when shooting the second depth image, and the door position in the second depth image, the real-world door position and door direction are obtained.

[0020] After determining the door position and door direction, the robot body is controlled to face the door direction and the drivable area from the current position to the door position is determined.

[0021] In some examples, there may be obstacles, such as seats, between the current position and the door position, and it is necessary to determine the drivable area in which the robot can move, which may be the floor area inside the vehicle.

[0022] S120: Control the robot body to move toward the vehicle door within the drivable area.

[0023] The drivable area is a connected area. Within the drivable area, the shortest path is planned with the current position as the starting point and the door position as the end point, and the robot body is controlled to move toward the door along the shortest path.

[0024] S130 , collecting a first depth image in the direction of the vehicle door, and performing door edge detection on the first depth image to obtain a position of a lower edge of the vehicle door.

[0025] When the robot body moves toward the car door, the depth camera on the robot body takes real-time photos in the direction of the moving front to obtain multiple first depth images. The door edge detection is performed on each first depth image, so as to timely determine the position of the lower edge of the car door (mainly the position of the lower edge of the car door), and control the robot to pause at the lower edge of the car door to avoid falling off the car.

[0026] S140: Control the robot body to get out of the vehicle according to the detected position of the lower edge of the vehicle door.

[0027] When the robot reaches the lower edge of the door, it can get off the car in one of the following two ways.

[0028] The first method is to jump out of the car directly: detect the height of the lower edge of the door from the ground, for example, by using a depth camera or radar. If the height is less than or equal to the maximum jumping height of the robot body, that is, the robot supports jumping at this height, then control the robot body to jump out of the car from the lower edge of the door.

[0029] Optionally, in addition to the height of the lower edge of the door from the ground, the robot also needs to detect the depth image to evaluate whether the getting-off area meets the following conditions: whether the ground is flat and whether there are obstacles on the ground.

[0030] The second method is to get off the vehicle by moving the disembarkation plate, which is overlapped between the lower edge of the door and the ground: the third depth image outside the door is collected by the depth camera; the edge of the disembarkation plate is detected on the third depth image to obtain the position of the edge of the disembarkation plate. The drivable area is combined with the position of the edge of the disembarkation plate to obtain a safety map; in the area corresponding to the safety map, the robot body is controlled to get off the vehicle along the disembarkation plate.

[0031] The key to the second method is to accurately identify the position of the edge of the board. Optionally, first, the third depth image is filtered to remove noise, such as using a median filter or a Gaussian filter for smoothing. Then, the depth data in the third depth image is normalized to a uniform range for subsequent processing, see the following formula: ; Wherein, D is each depth data in the third depth image. max is the maximum depth value in the third depth image. min is the minimum depth value in the third depth image. It is the normalized depth data.

[0032] Then, the Sobel operator is used to calculate the gradient of the third depth image to detect the edge. The Sobel operator can calculate the gradient of the image in the x and y directions: ; Among them, G represents the gradient amplitude, G x , G y Represents the Sobel operator.

[0033] In order to accurately locate the edge, non-maximum suppression (NMS) is used to retain the local maximum. Finally, the detected edge of the landing board is combined with the drivable area to generate the final safety map. Specifically, the edge of the landing board can be marked on the boundary of the drivable area to obtain a safety map. For safety maps, see Figure 2As shown in the figure, it is obtained by splicing the edge of the disembarkation board (red) and the drivable area (green). The safety map is projected into the real world to obtain the area where the robot can move. Then, autonomous disembarkation is achieved within the movable area.

[0034] In actual operation, as the robot moves, the third depth image and edge detection results need to be dynamically updated, and the detection parameters need to be adjusted to cope with environmental changes. A filter (such as a Kalman filter) can be used to smooth the detection results and track the edge position. This embodiment achieves accurate disembarkation by identifying the position of the edge of the disembarkation plate.

[0035] In order to achieve accurate exit, this embodiment combines the drivable area and edge detection technology. The depth camera detects the lower edge of the door and the edge of the exit panel, and integrates this edge information with the drivable area data, which can achieve a more accurate autonomous exit function and improve the safety and accuracy of exit.

[0036] Compared with the prior art, the embodiments of the present application also have the following technical effects: The robot on the vehicle senses and locates the environment, determines the drivable area toward the door and automatically moves to the door; then it detects the door position and controls the robot body to get off the vehicle, providing a complete solution for the robot to get off the vehicle autonomously. This embodiment ensures the safety of the robot's movement by detecting the drivable area and the lower edge of the door, enhances the robot's adaptability and autonomy in various terrains, and thus achieves higher operational reliability and safety. The solution provided by this embodiment mainly relies on the sensing device on the robot body to make control decisions, avoids dependence on external signals, and improves the independence and flexibility of the system.

[0037] The present application also provides a method for the robot to autonomously get off a vehicle, see Figure 3 , refine the door lower edge detection algorithm and the drivable area determination algorithm to improve detection accuracy, including the following operations: S210 , in response to the instruction to get off the vehicle, collecting point cloud data from the current position to the door position.

[0038] Through the laser radar on the robot body, point cloud data from the current position to the door position is collected. The point cloud data is composed of the three-dimensional coordinates of each point in the environment captured by the radar. It has high-precision and detailed spatial information. Each point cloud is represented by a coordinate point (x, y, z) in the laser radar coordinate system. The laser radar is the origin, z represents the vertical height of the point cloud in the laser radar, y represents the horizontal position of the point cloud in the laser radar, and x represents the vertical position of the point cloud in the laser radar.

[0039] S220 , performing ground segmentation on the point cloud data to obtain ground point cloud data, wherein the ground point cloud data constitutes the drivable area.

[0040] Some point cloud data are ground point clouds, and the robot can move. In this case, the ground point cloud data is used to form the drivable area. Some point cloud data are obstacle point clouds, and the robot cannot move. Therefore, it is necessary to segment the point cloud data into ground point cloud data and obstacle point cloud data. An optional implementation of ground segmentation is provided below, including S221 to S224.

[0041] S221. Rasterize the point cloud data to obtain a grid map, wherein each grid in the grid map includes joint height data of a plurality of point cloud data.

[0042] The point cloud data collected by the S210 is rasterized and converted into a 2.5D grid map, which divides the environment into regular grids and records the height information in each grid, so as to describe the ups and downs of the terrain. The purpose of rasterization is to convert complex point cloud data into a regular, easy-to-process grid format, making the environmental information more structured and simplified. Through rasterization, the point cloud data is divided into uniform grid units (i.e., grids), and each grid records specific environmental features (such as obstacles or ground), making subsequent map processing, analysis, and obstacle avoidance calculations more efficient and intuitive.

[0043] First, choose an appropriate grid size (usually d x and d y In this scheme, 0.05 is taken, that is, the size of each grid is a 5 cm square). The size of the grid affects the resolution of the final grid map. Mapped to a two-dimensional grid, the index of the point cloud position in the grid can be calculated by the following formula ( , ), used to normalize the point cloud: ; ; in, and is the minimum value of the x-coordinate and y-coordinate in the point cloud data, and i is the number of the point cloud.

[0044] Then, the z coordinate value of the point cloud is mapped to the corresponding two-dimensional grid. Each grid will contain the z values ​​of multiple point clouds, and the joint height data (such as the maximum height or average height) of each point cloud in the grid can be calculated to represent the terrain height of the grid. The final 2.5D grid map is as follows Figure 4 As shown, each color block is a grid, and different colors are used to represent different joint height data / terrain height data, recorded as .

[0045] S222: Map each grid in the grid map to a plurality of sector areas, each sector area includes a plurality of containers, and each container is mapped with comprehensive height data of a plurality of grids.

[0046] After obtaining the 2.5D grid image, it is necessary to perform ground segmentation. This is usually achieved by analyzing the height and position of the point cloud, with the goal of removing obstacle information and retaining ground information. In order to be applicable to application scenarios where the ground inside the car has a certain slope, this embodiment provides a method based on ground straight line fitting to achieve ground segmentation. First, see Figure 5 As shown, the circle is divided into multiple sector areas (i.e., N segments), the angle of each segment is Δα, and each segment is equally divided into M containers (bins), and each container is represented as segments_ (bin). A segment is used as an independent processing unit for straight line fitting.

[0047] Project each grid into segments_ (bin). That is, project the grid P ( , , ) is reduced to P'(d, ), where d is the position of the grid in the ground coordinate system.

[0048] S223. Determine the minimum comprehensive height information of each container in each sector area, and perform piecewise linear fitting on the minimum comprehensive height information of each container in each sector area to obtain a ground line.

[0049] Take the lowest point in each container, fit the ground line based on these lowest points within the fan-shaped area, and use this as a basis to determine whether the grid is a ground point, so as to achieve the segmentation of ground and non-ground. For example, fit a straight line Y = kd + b for each segment, where k refers to the slope of the ground and b is the intercept. If the ground has a slope, the fitted straight line is segmented, and different coefficients (k, b) represent grounds with different slopes.

[0050] S224, selecting a target grid whose comprehensive height data is lower than the ground line, and determining all point cloud data in the target grid as ground point cloud data.

[0051] For all grids, if the integrated height data is less than or equal to the threshold, the grid is determined to be the target grid and the point cloud data therein is the ground point; if the integrated height data is greater than the threshold, the point cloud data therein is the obstacle point.

[0052] like Figure 6As shown, the above method can set the target grid in the 2.5D grid map to gray, indicating that it is drivable. Set other grids to black, indicating that it is not drivable, to form a 2D occupancy grid map. The area where the gray grid is located is used as the drivable area for path planning. Exemplarily, the 2D occupancy grid map is passed to the robot controller, and the optimal path is calculated based on the drivable area in the 2D occupancy grid map to avoid obstacles and realize dynamic obstacle avoidance function. In this process, the controller will update the map data in real time to respond to changes in the environment and ensure that the robot can move safely.

[0053] In the obstacle avoidance function of this embodiment, in order to save computing power and resources, a grid map is used to represent the drivable area, providing a simple and fast debugging environment for the entire boarding and alighting function link.

[0054] S230: Control the robot body to move toward the vehicle door within the drivable area.

[0055] S240: Collect a first depth image in the direction of the vehicle door.

[0056] When the robot moves toward the door, the first depth image in the direction of the door is collected in real time by the depth camera to detect the position of the lower edge of the door in real time.

[0057] S250 , performing multi-scale sampling on the first depth image to obtain images to be detected with multiple resolutions.

[0058] Since the lower edge of the door is generally a straight line, this embodiment adopts an effective image straight line detection algorithm. Since the traditional Hough Transform algorithm has the defects of large computational complexity, high memory consumption, sensitivity to noise, difficulty in processing complex shapes and difficulty in parameter adjustment, this embodiment proposes a multi-scale Hough Transform algorithm to detect edge lines in the first depth image.

[0059] Optionally, the first depth image is continuously downsampled, and after each downsampling, the image resolution is reduced by a set ratio in the row and column directions to obtain multiple images to be detected with gradually reduced resolutions. The set ratio can be set according to demand, for example, 50%. For example, the first depth image is continuously downsampled using an image pyramid algorithm. An image pyramid is a multi-scale expression of an image, which is an effective but conceptually simple structure for interpreting an image at multiple resolutions. Each level of the image is a downsampled version of the image at the previous level. The formula can be expressed as: ; in, is the image at scale s, Q is the downsampling filter, is the image at the previous scale.

[0060] S260 , performing straight line detection on the image to be detected at each resolution to obtain edge lines of multiple resolutions.

[0061] In each image to be detected at each scale s, edge detection is first performed, such as the Canny operator, Sobel operator, etc. After edge detection, a binary image is obtained, in which the edge pixel value is 1 and the non-edge pixel value is 0. The Hough transform algorithm is used to accumulate each detected edge pixel in the parameter space. For line detection, the parameter space is generally represented by polar coordinates, including two parameters: polar diameter (distance) and polar angle (angle). For each edge pixel, according to its position in the image to be detected, the corresponding straight line in the parameter space is calculated and accumulated in the parameter space.

[0062] Find the points with the largest cumulative value in the parameter space. These points represent the parameters of the possible straight lines. Decode the points found in the parameter space back to the image space to obtain the parameters of the detected straight line. In this way, the edge lines of the image to be detected at each resolution can be obtained, that is, multi-scale edge lines. Assume that the parameters of the edge lines include the slope and the intercept .

[0063] S270: Obtain the position of the lower edge of the vehicle door in the first depth image according to the edge lines of multiple resolutions.

[0064] Assume that there are f resolutions, i.e., f images to be detected. If one straight line is detected for each image to be detected, then a total of f straight lines are detected. The slopes of the f straight lines are averaged, and the intercepts are averaged to obtain a final straight line, which is used to indicate the position of the lower edge of the car door.

[0065] S280: Control the robot body to get out of the vehicle according to the detected position of the lower edge of the vehicle door.

[0066] In an actual application scenario, when the robot body is getting off the car, it needs to rely on the robot's balance ability and accurate recognition of the lower edge of the door to avoid falling or collision, ensuring the safety and efficiency of the getting off process. Take the robot dog as an example. The robot dog is a robot that can simulate the movements and behaviors of a real dog. It has the ability to move autonomously and stabilize itself. The robot dog is equipped with various sensors, including but not limited to gyroscopes, accelerometers, and tilt sensors, which can monitor the posture and motion state of the robot dog in real time. Through these sensors, the robot dog can sense its own tilt angle and direction of movement, and make corresponding adjustments. When the robot dog detects that it is tilting or losing balance, the control algorithm will respond immediately and maintain balance by adjusting the robot dog's joints and motion trajectory. In addition to intelligent perception and control algorithms, the structural design of the robot dog also plays an important role in its self-stabilization. The robot dog can be wheeled or legged (see Appendix Figure 7 ) design, which can provide better stability and flexibility.

[0067] When the robot reaches the ground, it can use the depth camera and lidar on the robot to identify the designated location, accurately locate and navigate to the designated location. In addition, the robot can also activate the autonomous charging function after reaching the ground. After reaching the designated location, the robot can automatically navigate to the charging station for energy replenishment and prepare for the next mission.

[0068] In this embodiment, the point cloud data is first rasterized to convert the three-dimensional point cloud data into a two-dimensional occupancy grid map. This process significantly improves the processing efficiency by mapping the point cloud data onto the grid, and can quickly generate a 2D occupancy grid map that can be directly used for path planning and regulation. In order to cope with ramps and irregular terrain, this solution proposes an obstacle avoidance algorithm based on ground segmentation. By segmenting the ground, the algorithm can adapt to different terrain changes, thereby enhancing the robot's motion scene capabilities. The multi-scale Hough transform in this embodiment is an extension of the Hough transform, which can detect geometric shapes of different sizes; by applying the Hough transform at multiple scales to improve detection flexibility and applicability, the accuracy of edge detection is improved.

[0069] In summary of the above embodiments, the present application greatly improves the overall efficiency and adaptability of the robot through point cloud rasterization processing, multi-scale Hough transform algorithm and optimization of autonomous getting off function.

[0070] The present application also provides a robot, see Figure 7, including: a controller (built into the robot body) and a robot body connected to the controller. This embodiment does not limit the appearance of the robot, such as the shape of a robot dog. The robot body at least includes a forward drive device and a steering device, which moves forward or turns in the car under the control command of the controller. A sensing device is installed on the robot body for collecting images in the direction of the door. The controller is used to determine the drivable area from the current position to the door position in response to the get-off instruction; within the drivable area, control the robot body to move toward the door; collect a first depth image in the direction of the door, and perform door edge detection on the image to obtain the position of the lower edge of the door; according to the detected position of the lower edge of the door, control the robot body to get off the car.

[0071] The controller is also used to execute the method for the robot to autonomously get off the vehicle provided in any of the above embodiments, and has corresponding technical effects, which will not be repeated here.

[0072] Optional, continue to see Figure 7 The sensing device includes a laser radar and a depth camera. The laser radar is used to collect point cloud data from the current position to the door position; the depth camera is used to collect a second depth image around the robot body in response to the command to get off the vehicle; a first depth image around the robot body; and a third depth image outside the door. The robot is equipped with a battery for charging the sensing device, the forward drive device, and the steering device.

[0073] Optionally, a normal camera is installed on the robot body, which is used for pedestrian following and face recognition. In order to see most of the features of the human body at close range, both the normal camera and the laser radar are tilted upward at a certain angle. After actual measurement, the normal camera and the laser radar have almost no blind spots at close range.

[0074] Optionally, a microphone and a voice processing device are also installed on the robot body to collect the get-off instructions issued by the vehicle and perform semantic analysis.

[0075] The controller, microphone, voice processing device, perception device, forward drive device and steering device use the ROS2 (Robot Operating System 2) communication mechanism. ROS 2 is an open source platform for robot development, which provides a series of tools and libraries for building robot applications. ROS 2 is designed with more real-time, safety and reliability factors in mind, so it is suitable for a wider range of application scenarios and hardware platforms.

[0076] The embodiment of the present application also provides a robot autonomous getting-off system, comprising a robot and a vehicle. The robot is located in the vehicle. The vehicle is used to issue a get-off command to the robot; after issuing the get-off command, the vehicle detects whether the door is open; if not, the vehicle notifies a person to open the door, or controls the door to open automatically.

[0077] In summary of the above embodiments, this application uses the robot's own laser radar and depth camera through the perception device to perform efficient environmental perception and self-positioning, avoiding dependence on external signals. The point cloud rasterization process is used to improve the flexibility and processing efficiency of the system. The innovative ground segmentation algorithm adapts to a variety of terrains, and the autonomous disembarkation function combined with the drivable area and edge detection improves the accuracy and safety of disembarkation. These technical points together improve the overall efficiency and adaptability of the robot in complex environments.

[0078] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for a robot to autonomously get off a vehicle, characterized in that: include: In response to the vehicle exit instruction, determining a drivable area from a current position to a vehicle door position; In the drivable area, controlling the robot body to move toward the vehicle door; Collecting a first depth image in the direction of the vehicle door, and performing door edge detection on the first depth image to obtain the position of the lower edge of the vehicle door; According to the detected position of the lower edge of the door, the robot body is controlled to get off the vehicle.

2. The method according to claim 1, characterized in that Performing door edge detection on the first depth image to obtain a position of a lower edge of the door includes: Performing multi-scale sampling on the first depth image to obtain images to be detected with multiple resolutions; Perform straight line detection on the image to be detected at each resolution to obtain edge lines of multiple resolutions; The position of the lower edge of the vehicle door in the first depth image is obtained according to the edge lines of the multiple resolutions.

3. The method according to claim 2, characterized in that The first depth image is subjected to multi-scale sampling to obtain images to be detected with multiple resolutions, including: The first depth image is continuously downsampled, and after each downsampling, the image resolution is reduced by a set ratio in the row and column directions to obtain a plurality of to-be-detected images with gradually reduced resolutions.

4. The method according to claim 1, characterized in that The step of determining a drivable area from a current position to a door position in response to a get-off instruction includes: In response to the get-off instruction, collecting a second depth image around the robot body; Determining a door direction and a door position according to the depth data in the second depth image; The robot body is controlled to face the vehicle door direction, and a drivable area from the current position to the vehicle door position is determined.

5. The method according to claim 1, characterized in that: The determining of the drivable area from the current position to the door position includes: Collect point cloud data from the current position to the door position; The point cloud data is subjected to ground segmentation to obtain ground point cloud data, and the ground point cloud data constitutes the drivable area.

6. The method according to claim 5, characterized in that Performing ground segmentation on the point cloud data to obtain ground point cloud data also includes: Performing rasterization processing on the point cloud data to obtain a raster map, wherein each grid in the raster map includes joint height data of a plurality of point cloud data; Mapping each grid in the grid map into a plurality of sector areas, each sector area including a plurality of containers, each container being mapped with comprehensive height data of a plurality of grids; Determine the minimum comprehensive height information of each container in each sector area, and perform piecewise linear fitting on the minimum comprehensive height information of each container in each sector area to obtain a ground line; A target grid whose integrated height data is lower than the ground line is selected, and the point cloud data in the target grid is determined as ground point cloud data.

7. The method according to claim 1, characterized in that According to the detected position of the lower edge of the door, the robot body is controlled to get off the vehicle, including the following first process or second process: The first process: detecting the height of the lower edge of the door from the ground; If the height is less than or equal to the maximum jumping height of the robot body, controlling the robot body to jump off the vehicle from the lower edge of the door; Second process: collecting a third depth image outside the vehicle door, and performing alighting board edge detection on the third depth image to obtain the position of the alighting board edge; combining the drivable area with the position of the alighting board edge to obtain a safety map; In the area corresponding to the safety map, the robot body is controlled to get off the vehicle along the getting off board.

8. A robot, characterized in that: include: A controller and a robot body connected to the controller; The robot body is equipped with a sensing device for collecting images in the direction of the vehicle door; The controller is used to respond to a get-off instruction and determine a drivable area from a current position to a door position; within the drivable area, control the robot body to move toward the door; collect a first depth image in the direction of the door, and perform door edge detection on the image to obtain the position of the lower edge of the door; and control the robot body to get off the vehicle according to the detected position of the lower edge of the door.

9. The robot according to claim 8, characterized in that: The sensing device comprises: LiDAR, used to collect point cloud data from the current position to the door position; The depth camera is used to collect a second depth image around the robot body in response to a get-off instruction; a first depth image around the robot body; and a third depth image outside the vehicle door.

10. A robot autonomous disembarking system, characterized in that: include: The robot and vehicle according to claim 8 or 9; The robot is located in the vehicle; The vehicle is used to issue a get-off instruction to the robot.

Images