Robot charging pile docking method and system based on vision

Abstract

The invention discloses a robot charging pile docking method and system based on vision. The docking method comprises the following steps that an edge end computing platform and a visible light camera are carried on a charging gun; when the edge end computing platform receives a charging request of a robot control system, acquiring the latest acquired image, detecting the position of a preset icon on the robot in the image, and judging whether the position meets the requirement or not; if the requirement is met, calculating the length-width ratio of a preset icon in the image view; s4, when the length-width ratio does not meet the requirement, the posture of the charging gun is adjusted; when it is detected that the posture of the charging gun is consistent with the current posture of the robot, the distance between the charging gun of the charging pile and a charging port in the robot is calculated. The invention further comprises a robot charging pile docking system based on vision. The system is low in cost and accurate in positioning, and through real-time monitoring and positioning of the robot and the charging pile, it is effectively ensured that the charging gun is accurately aligned with the charging port and inserted into the charging port for charging.

Description

Technical Field

[0001] The present invention relates to the technical field of robot charging, and in particular to a vision-based robot charging pile docking method and system. Background Art

[0002] Currently, when robots are charging and locating, they use the method of using the robot's front laser radar to collect information about the reflector next to the charging pile to determine the location of the charging port. However, this method often causes an error of about 10 cm when the robot is aligning the charging port due to position loss or odometer error.

[0003] In order to solve the problem of inaccurate positioning, the application number 202411004762.9 "Vehicle Charging Port Positioning Method Based on Visual Recognition" proposes a positioning method that integrates laser radar and visible light vision. The visual system of this method includes a laser transmitter and a camera. The laser transmitter emits a laser to aim at the vehicle charging port, and the visual system shoots the charging port, and generates a point cloud map through image intelligent analysis, and calculates the accurate position coordinate information of the vehicle charging port: S1: Scanning and collecting to generate a point cloud map; S2: Image preprocessing to generate a model: Using the acquired point cloud data, rasterizing after noise point processing, generating a digital surface model, rasterizing by constructing an irregular triangulated network model, obtaining digital elevation model data, or converting it into contour data; S3: Coordinate transformation: Coordinate transformation is performed in sequence, pixel coordinates-image coordinates-camera coordinates-robot coordinates, to determine the charging port position. However, the algorithm used in this patent document requires additional laser radar equipment, which is costly for ordinary robots, and requires multiple coordinate transformations, resulting in a very complex algorithm. Summary of the invention

[0004] The purpose of the present invention is to overcome the above-mentioned deficiencies in the prior art and to provide a vision-based robot charging pile docking method and system with low cost, good robustness and high positioning accuracy.

[0005] The technical solution of the present invention is: A vision-based robot charging pile docking method of the present invention comprises the following steps: S1: The edge computing platform and the visible light camera are mounted on the charging gun. After the edge computing platform is powered on, the image acquisition thread is enabled to cyclically collect image data. S2: When the robot needs to charge, it navigates to the vicinity of the charging pile according to the location of the charging pile in the map, obtains the reflector information of the charging pile according to the laser radar on the robot, completes the initial positioning with the charging pile, and the robot control system initiates a charging request to the edge computing platform; S3: When the edge computing platform receives a charging request from the robot control system, it obtains the latest captured image and detects the position of the preset icon on the robot in the image. If the preset icon is not detected or the position of the preset icon does not meet the moving range requirement of the charging gun on the charging pile, the edge computing platform feeds back control logic information to the robot; S4: Detect the number of length and width pixels of the preset icon in the image, calculate the aspect ratio of the preset icon in the image field of view, and when the aspect ratio exceeds a preset threshold, determine that the current robot posture is offset; S5: When the aspect ratio of the preset icon exceeds the preset threshold, obtaining current robot posture information; S6: Calculate the rotation matrix relative to the optical center direction of the visible light camera according to the current robot posture information, and adjust the posture of the charging gun of the charging pile so that the posture of the charging gun of the charging pile is consistent with the current robot posture; S7: When it is detected that the posture of the charging gun is consistent with the current posture of the robot, the distance between the charging gun of the charging pile and the charging port on the robot is calculated through the pixel size of the preset icon in the camera field of view and the camera focal length information, and the precise position coordinate information of the charging port is obtained to guide the translation of the charging gun of the charging pile.

[0006] Furthermore, the preset icon is a circular icon.

[0007] Further, in S3, the control logic information fed back by the edge computing platform includes: returning 0, indicating that the current charging port position of the robot meets the requirements; returning 1, indicating that the preset icon is not detected and the robot needs to reposition the charging pile; returning 2, indicating that the current robot needs to move left; returning 3, indicating that the current robot needs to move right.

[0008] Further, in S7, the distance between the charging gun of the charging pile and the charging port on the robot is obtained by the following formula: In the formula, p is the pixel size of the preset icon in the camera field of view, f is the focal length of the visible light camera, d is the actual length of the preset icon diameter, z The actual distance from the optical center of the visible light camera to the preset icon, in units of d Consistent.

[0009] Further, in S5, when the aspect ratio of the preset icon exceeds the preset threshold, the edge computing platform uses a monocular 3D target detection model to detect the 3D frame of the robot and obtains the current robot posture information from the 3D frame information.

[0010] Furthermore, the method further includes step S8: detecting whether the charging current is obtained. If the charging current is obtained, it indicates that the robot is successfully docked with the charging gun and charging has started; if the charging current is not obtained, steps S3 to S7 are repeated.

[0011] A vision-based robot charging pile docking system of the present invention includes: A visible light camera, mounted on the charging gun, is used to collect image data of the robot during the charging process, including the collection of preset icons on the robot; The edge computing platform is mounted on the charging gun and is used to cyclically collect image data captured by the visible light camera; it is used to obtain the position of the robot preset icon in the image according to the charging request sent by the robot control system, and when the preset icon is not detected or the position of the preset icon does not meet the moving range requirement of the charging gun on the charging pile, the control logic information is fed back to the robot control system; it is used to analyze the preset icon in the image, extract its length and width pixel values, and calculate the aspect ratio of the preset icon in the image field of view; it is used to evaluate whether the aspect ratio of the preset icon exceeds the set threshold. If it exceeds the threshold, the target detection model is used to perform robot 3D bounding box detection, and extract the current robot posture information from the 3D bounding box; it is used to calculate the rotation matrix relative to the optical center direction of the visible light camera according to the current robot posture information, and send a charging gun adjustment instruction to the charging gun control system according to the rotation matrix; it is also used to calculate the distance between the charging gun of the charging pile and the charging port of the robot based on the pixel size of the preset icon in the camera field of view and the focal length of the camera when it is detected that the posture of the charging gun is consistent with the current robot posture, and send a charging gun translation instruction to the charging gun control system according to the distance; The robot control system is used to guide the robot to navigate to the vicinity of the charging pile according to the location of the charging pile in the map; to control the robot's movements according to the reflector information of the charging pile obtained by the laser radar to complete the initial positioning with the charging pile; to initiate a charging request to the edge computing platform and adjust the robot's movement according to the control logic information fed back by the edge computing platform; The charging gun control system is used to control the charging gun movement according to the charging gun adjustment instructions sent by the edge computing platform, so that the charging gun posture is consistent with the current robot posture; it is also used to control the charging gun to move to the target position according to the charging gun translation instructions sent by the edge computing platform.

[0012] Furthermore, the charging gun of the charging pile can move in the front-rear direction and can adjust the angle.

[0013] Furthermore, the edge computing platform uses a similar triangular relationship to calculate the distance between the charging gun of the charging pile and the charging port on the robot.

[0014] Furthermore, the robot control system is also used to detect whether the charging current is obtained and send corresponding instructions to the edge computing platform.

[0015] The beneficial effects of the present invention are as follows: on the one hand, by installing both the edge computing platform and the visible light camera on the charging gun, the delay in data transmission is greatly reduced, and the real-time performance and response speed of the system are improved; and by combining a preset circular icon with a visible light camera, it has higher reliability and robustness, lower cost, and the overall algorithm is simple; on the other hand, by adjusting the posture of the robot and the charging gun, it can ensure that the docking between the charging gun and the robot is more precise. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of a method flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figure 1 As shown: A vision-based robot charging pile docking method includes the following steps: S101: An edge computing platform and a visible light camera are mounted on a charging gun. After power is turned on, the edge computing platform enables an image acquisition thread to cyclically collect image data.

[0019] In the above steps, on the one hand, by mounting both the edge computing platform and the visible light camera on the charging gun, image acquisition and processing are performed directly on the charging gun, so that the edge computing platform enables the image acquisition thread to cyclically acquire image data, which greatly reduces the delay in data transmission and improves the real-time performance and response speed of the system. If the camera is installed on the robot, it is usually necessary to transmit the image data to the robot control system for processing, which increases the complexity of data transmission and processing; on the other hand, the visible light camera is directly installed on the charging gun, which can more accurately detect the position of the charging port, reduce the position error caused by the movement of the robot, ensure that the charging gun can accurately align with the charging port, and reduce the camera position offset caused by the movement or vibration of the robot, which improves the reliability of the system in complex environments.

[0020] S102: When the robot needs to charge, it navigates to the vicinity of the charging pile according to the location of the charging pile in the map, and then obtains the reflector information of the charging pile according to the laser radar on the robot to complete the initial positioning with the charging pile. The robot control system initiates a charging request to the edge computing platform.

[0021] Specifically, the present invention does not abandon the technical solution of using the reflector of the charging pile for positioning, but integrates it into the solution of the present invention. By using a laser radar to obtain the information of the reflector of the charging pile for initial positioning, the accuracy and reliability of the initial positioning are improved, ensuring that the robot can quickly and accurately navigate to the vicinity of the charging pile. This fusion technology not only improves the accuracy of the initial positioning, but also greatly reduces the error range that needs to be processed by the subsequent precise positioning algorithm, simplifies the complexity of precise positioning, and improves the efficiency of the overall system.

[0022] S103: When the edge computing platform receives a charging request from the robot control system, it obtains the latest captured image and detects the position of the circular icon pre-installed on the robot in the image. If the preset circular icon is not detected or the position of the preset circular icon does not meet the moving range requirement of the charging gun on the charging pile (the range value can be obtained based on the actual physical equipment test), the edge computing platform feeds back the controller interface information to the robot: return 0, indicating that the current charging port position on the robot meets the requirements; return 1, indicating that the preset circular icon is not detected and the robot needs to reposition the charging pile; return 2, indicating that the current robot needs to move left; return 3, indicating that the current robot needs to move right.

[0023] Specifically, the present embodiment pre-sets a circular icon at the target position of the robot. The circular icon can be an icon that is clearly distinguished from the background, such as placing a black circular sign on a white robot; or there is an obvious border around the circular icon, and the border color contrasts with the background and the circle itself. The reason for designing it as a circle is that: on the one hand, circles are relatively rare in the natural environment, have unique shapes, and have a low probability of false detection; on the other hand, the circle detection algorithm is relatively simple and robust, and can be recognized more effectively. The prior art usually uses a camera combined with a laser radar device, or a camera with an emitted light source to locate the robot, which not only increases the cost and the calculation difficulty of the algorithm, but also easily leads to detection failure due to light source failure or changes in ambient lighting; the present invention uses a preset circular icon combined with a visible light camera, which has higher reliability and robustness, and has lower cost.

[0024] Furthermore, this embodiment returns the controller interface information to the robot according to the preset circular icon position, and the feedback control logic is simple and clear, easy to implement and debug; that is, it has the following advantages: First, the return values ​​0, 1, 2, and 3 provide clear instructions, and the robot control system can quickly understand and execute these instructions without complex decision-making logic; second, the edge computing platform enables the image acquisition thread to cyclically collect image data after power-on, and can detect the position of the preset circular icon on the robot in real time, and immediately feedback control information, which greatly improves the response speed, especially in complex or dynamically changing environments. The real-time feedback and rapid adjustment capabilities enable the system to better adapt to environmental changes and improve robustness; third, returns 2 and 3 indicate that the robot needs to move left or right, respectively, providing precise adjustment directions to ensure that the robot can accurately align with the charging port; fourth, the edge computing platform is responsible for image processing and position detection, and the robot control system only needs to perform simple actions according to the return value, which simplifies the algorithm complexity of the overall system, so that the robot control system does not need to perform complex image processing and position detection, reduces the demand for computing resources, and improves the overall performance of the system.

[0025] S104: By detecting the number of length and width pixels of the preset circular icon in the image, the aspect ratio of the preset circular icon in the image field of view is calculated, and whether the current robot posture is offset is confirmed based on the aspect ratio: when the aspect ratio exceeds the threshold (obtained through empirical testing and related to the selected visible light camera), it is determined that the current robot posture is offset.

[0026] Specifically, when the edge computing platform feeds back the controller interface information to the robot, how to determine the distance of the robot's left or right movement is to determine the robot's posture offset by calculating the aspect ratio of the preset circular icon in the image field of view, which has higher robustness and higher accuracy than judging by the pixel size of the marker. That is, on the one hand, the aspect ratio is a relatively stable feature. Even in the case of lighting changes, background interference or image noise, the change in the aspect ratio can still accurately reflect the deformation degree of the circular icon, thereby judging the robot's posture offset; and the aspect ratio can more accurately reflect the deformation degree of the circular icon in the image, thereby more accurately judging the robot's posture offset; while the pixel size is easily affected by factors such as ambient lighting, camera focal length changes, background interference, etc., resulting in false detection and misjudgment; and the pixel size can only provide the size information of the marker, and cannot directly reflect the posture offset. On the other hand, the calculation of the aspect ratio is relatively simple, and it is only necessary to calculate its aspect ratio after detecting the circular icon.

[0027] S105: When the aspect ratio of the preset circular icon exceeds the preset threshold, a monocular 3D target detection model is used to detect the robot 3D frame, and current robot posture information is obtained from the 3D frame information.

[0028] Specifically, if the aspect ratio of the preset circular icon exceeds the preset threshold, it means that the circular icon has been deformed in the image, which means that the robot has a posture deviation. At this time, the edge computing platform will use a monocular 3D target detection model to detect the robot's 3D frame information, including pitch angle, yaw angle, roll angle, etc. For example: the 3D frame information shows that the robot's yaw angle is 10°, the roll angle is 5°, etc. Using this information can help the system more accurately determine the robot's posture deviation and make corresponding adjustments, thereby improving the overall robustness of the system.

[0029] S106: Calculate the rotation matrix relative to the optical center direction of the visible light camera according to the current robot posture information, and adjust the charging gun posture of the charging pile so that the charging gun posture of the charging pile is consistent with the current robot posture.

[0030] Specifically, this embodiment obtains the current robot posture information through a monocular 3D target detection model, and then adjusts the charging gun posture of the charging pile according to the posture information. On the one hand, it can more accurately align the robot charging port, taking into account not only the robot posture, but also the charging gun posture, ensuring the consistency of the two during docking, thereby greatly improving the docking success rate and accuracy; on the other hand, in a dynamic environment, the robot's posture may change due to various factors (such as uneven ground). By adjusting the charging gun posture in real time, it can better adapt to these changes and ensure the stability and reliability of the charging process. If only the robot posture is adjusted, the robot may need to move and adjust significantly, which is not only time-consuming, but may also increase the robot's energy consumption; by adjusting the charging gun posture, accurate docking can be achieved without increasing the robot's adjustment burden, thereby improving charging efficiency.

[0031] S107: When it is detected that the posture of the charging gun is consistent with the current posture of the robot, the distance between the charging gun of the charging pile and the charging port on the robot is calculated using the similar triangular relationship by presetting the pixel size of the circular icon in the camera field of view and the camera focal length information, and the precise position coordinate information of the charging port is obtained to guide the translation of the charging gun of the charging pile.

[0032] Specifically, the distance between the charging gun of the charging pile and the charging port on the robot is obtained by the following formula: In the formula, p The pixel size of the preset circular icon in the camera field of view. f is the focal length of the visible light camera, d is the actual length of the preset circular icon diameter, z The actual distance from the optical center of the visible light camera to the preset circular icon, in units of d Consistent.

[0033] The charging gun of the charging pile is controlled to move horizontally according to the distance z, that is, the distance z is moved horizontally.

[0034] S108: Check whether the charging current is obtained. If it is obtained, it means that the robot and the charging gun are successfully docked and charging has started; if it is not obtained, repeat steps S103 to S107.

[0035] According to the above method, the vision-based robot charging pile docking system of the present invention includes: A visible light camera, mounted on the charging gun, is used to collect image data of the robot during the charging process, including the collection of preset icons on the robot; The edge computing platform is mounted on the charging gun and is used to cyclically collect image data captured by the visible light camera; it is used to obtain the position of the robot preset icon in the image according to the charging request sent by the robot control system, and when the preset icon is not detected or the position of the preset icon does not meet the moving range requirement of the charging gun on the charging pile, the control logic information is fed back to the robot control system; it is used to analyze the preset icon in the image, extract its length and width pixel values, and calculate the aspect ratio of the preset icon in the image field of view; it is used to evaluate whether the aspect ratio of the preset icon exceeds the set threshold. If it exceeds the threshold, the target detection model is used to perform robot 3D bounding box detection, and extract the current robot posture information from the 3D bounding box; it is used to calculate the rotation matrix relative to the optical center direction of the visible light camera according to the current robot posture information, and send a charging gun adjustment instruction to the charging gun control system according to the rotation matrix; it is also used to calculate the distance between the charging gun of the charging pile and the charging port of the robot based on the pixel size of the preset icon in the camera field of view and the focal length of the camera when it is detected that the posture of the charging gun is consistent with the current robot posture, and send a charging gun translation instruction to the charging gun control system according to the distance; The robot control system is used to guide the robot to navigate to the vicinity of the charging pile according to the location of the charging pile in the map; to control the robot's movements according to the reflector information of the charging pile obtained by the laser radar to complete the initial positioning with the charging pile; to initiate a charging request to the edge computing platform and adjust the robot's movement according to the control logic information fed back by the edge computing platform; and to detect whether the charging current is obtained and send an instruction to obtain or not obtain to the edge computing platform; The charging gun control system is used to control the charging gun movement according to the charging gun adjustment instructions sent by the edge computing platform, so that the charging gun posture is consistent with the current robot posture; it is also used to control the charging gun to move to the target position according to the charging gun translation instructions sent by the edge computing platform.

[0036] This embodiment can ensure more accurate docking between the charging gun and the robot by adjusting the posture of the robot and the charging gun. If the robot or the charging gun is controlled alone, it may not be able to fully adapt to environmental changes or the precise position of the charging pile, resulting in unstable or failed docking. Moreover, if only the robot is controlled to dock with the charging gun, the robot needs to locate the position of the charging gun very accurately and make high-precision adjustments. By controlling the action of the charging gun, the requirements for the positioning accuracy of the robot can be reduced, so that the robot can slightly deviate from the ideal position, and the charging gun adjusts the position to complete the precise docking. On the other hand, the robot may face different charging pile positions or environmental factors. By collaboratively adjusting the posture of the robot and the charging gun, it can respond to various environmental changes more flexibly, thereby improving the adaptability and robustness of the system. Furthermore, by adjusting the charging gun and the robot, the charging docking process can be completed more quickly, reducing waiting time and improving the overall work efficiency of the system.

[0037] In this embodiment, the charging gun of the charging pile has the ability to move forward and backward and rotate left and right at a certain angle. For example, if a rail-type mobile charging pile is used, the charging gun can move forward and backward through the rail system and rotate left and right through the rotating device, and its movement and rotation are controlled by the charging gun control system.

[0038] In this embodiment, a fill light may be provided to increase the brightness of the image captured by the visible light camera to make the image clearer.

[0039] To summarize, on the one hand, the present invention greatly reduces the delay in data transmission and improves the real-time performance and response speed of the system by installing both the edge computing platform and the visible light camera on the charging gun; and by combining a preset circular icon with a visible light camera, it has higher reliability and robustness, lower cost, and a simple overall algorithm; on the other hand, by adjusting the posture of the robot and the charging gun, it can ensure that the docking between the charging gun and the robot is more precise.

Claims

1. A vision-based robot charging pile docking method, characterized in that: The following steps are involved: S1: The edge computing platform and the visible light camera are mounted on the charging gun. After the edge computing platform is powered on, the image acquisition thread is enabled to cyclically collect image data. S2: When the robot needs to charge, it navigates to the vicinity of the charging pile according to the location of the charging pile in the map, obtains the reflector information of the charging pile according to the laser radar on the robot, completes the initial positioning with the charging pile, and the robot control system initiates a charging request to the edge computing platform; S3: When the edge computing platform receives a charging request from the robot control system, it obtains the latest captured image and detects the position of the preset icon on the robot in the image. If the preset icon is not detected or the position of the preset icon does not meet the moving range requirement of the charging gun on the charging pile, the edge computing platform feeds back control logic information to the robot; S4: Detect the number of length and width pixels of the preset icon in the image, calculate the aspect ratio of the preset icon in the image field of view, and when the aspect ratio exceeds a preset threshold, determine that the current robot posture is offset; S5: When the aspect ratio of the preset icon exceeds the preset threshold, obtaining current robot posture information; S6: Calculate the rotation matrix relative to the optical center direction of the visible light camera according to the current robot posture information, and adjust the posture of the charging gun of the charging pile so that the posture of the charging gun of the charging pile is consistent with the current robot posture; S7: When it is detected that the posture of the charging gun is consistent with the current posture of the robot, the distance between the charging gun of the charging pile and the charging port on the robot is calculated through the pixel size of the preset icon in the camera field of view and the camera focal length information, and the precise position coordinate information of the charging port is obtained to guide the translation of the charging gun of the charging pile.

2. The vision-based robot charging pile docking method according to claim 1, characterized in that: The preset icon is a circular icon.

3. The vision-based robot charging pile docking method according to claim 1 or 2, characterized in that: In S3, the control logic information fed back by the edge computing platform includes: returning 0, indicating that the current charging port position of the robot meets the requirements; returning 1, indicating that the preset icon is not detected and the robot needs to reposition the charging pile; returning 2, indicating that the current robot needs to move left; returning 3, indicating that the current robot needs to move right.

4. The vision-based robot charging pile docking method according to claim 1 or 2, characterized in that: In S7, the distance between the charging gun of the charging pile and the charging port on the robot is obtained by the following formula: In the formula, p is the pixel size of the preset icon in the camera field of view, f is the focal length of the visible light camera, d is the actual length of the preset icon diameter, z The actual distance from the optical center of the visible light camera to the preset icon, in units of d Consistent.

5. The vision-based robot charging pile docking method according to claim 1 or 2, characterized in that: In S5, when the aspect ratio of the preset icon exceeds the preset threshold, the edge computing platform uses a monocular 3D target detection model to detect the 3D frame of the robot and obtains the current robot posture information from the 3D frame information.

6. The vision-based robot charging pile docking method according to claim 1 or 2, characterized in that: The method also includes step S8: detecting whether the charging current is obtained. If the charging current is obtained, it indicates that the robot is successfully docked with the charging gun and charging has started; if the charging current is not obtained, steps S3 to S7 are repeated.

7. A vision-based robot charging pile docking system, characterized in that: include: A visible light camera, mounted on the charging gun, is used to collect image data of the robot during the charging process, including the collection of preset icons on the robot; The edge computing platform is mounted on the charging gun and is used to cyclically collect image data captured by the visible light camera; Used to obtain the position of the robot preset icon in the image according to the charging request sent by the robot control system, and feedback control logic information to the robot control system when the preset icon is not detected or the position of the preset icon does not meet the moving range requirements of the charging gun on the charging pile; Used to analyze the preset icon in the image, extract its length and width pixel values ​​and calculate the aspect ratio of the preset icon in the image field of view; It is used to evaluate whether the aspect ratio of the preset icon exceeds the set threshold. If it exceeds the threshold, the target detection model is used to detect the robot's 3D bounding box and extract the current robot's posture information from the 3D bounding box. It is used to calculate the rotation matrix relative to the optical center direction of the visible light camera based on the current robot's posture information, and send a charging gun adjustment instruction to the charging gun control system based on the rotation matrix. It is also used to calculate the distance between the charging gun of the charging pile and the charging port of the robot based on the pixel size of the preset icon in the camera field of view and the focal length of the camera when it is detected that the posture of the charging gun is consistent with the current posture of the robot, and send a charging gun translation instruction to the charging gun control system according to the distance; The robot control system is used to guide the robot to navigate to the vicinity of the charging pile according to the location of the charging pile in the map; to control the robot's movements according to the reflector information of the charging pile obtained by the laser radar to complete the initial positioning with the charging pile; to initiate a charging request to the edge computing platform and adjust the robot's movement according to the control logic information fed back by the edge computing platform; The charging gun control system is used to control the charging gun movement according to the charging gun adjustment instructions sent by the edge computing platform, so that the charging gun posture is consistent with the current robot posture; it is also used to control the charging gun to move to the target position according to the charging gun translation instructions sent by the edge computing platform.

8. The vision-based robot charging pile docking system according to claim 7, characterized in that: The charging gun of the charging pile can move in the front-rear direction and can adjust the angle.

9. The vision-based robot charging pile docking system according to claim 7 or 8, characterized in that: The edge computing platform uses similar triangular relationships to calculate the distance between the charging gun of the charging pile and the charging port on the robot.

10. The vision-based robot charging pile docking system according to claim 7 or 8, characterized in that: The robot control system is also used to detect whether the charging current is obtained and send corresponding instructions to the edge computing platform.

Images