Visual guidance based charging pile adaptive alignment and positioning method

Abstract

The application relates to the technical field of vehicle charging, in particular to a charging pile self-adaptive alignment and positioning system and method based on visual guidance. Vehicle information of multiple target vehicles of different models is collected, and then the pretreated information is input into a positioning model; the positioning model is used to determine the vehicle model based on the vehicle information, and the relative pose deviation between a charging plug and a charging port of the target vehicle is calculated in combination with 3D point cloud and the charging port position, so that the charging plug is inserted into the charging port; finally, real-time vehicle information is input into the trained positioning model to complete the charging of the real-time vehicle. The application combines the 3D point cloud technology and the visual guidance to calculate the accurate relative pose deviation, and implements the closed-loop feedback movement strategy based on the coordinate difference value, realizes the accurate positioning of the charging plug at the sub-centimeter level or even the millimeter level, and significantly improves the success rate of the charging process and the reliability of the whole system.

Description

Technical Field

[0001] This invention relates to the field of vehicle charging technology, and more specifically to a vision-guided adaptive alignment and positioning system and method for charging piles. Background Technology

[0002] With the popularization of new energy vehicles and the development of autonomous driving technology, the demand for automation and intelligence in the charging process is becoming increasingly urgent. Currently, the charging process still requires manual cable plugging and unplugging.

[0003] Chinese patent CN108177542A discloses a charging method for a self-adjusting charging pile. The charging method includes the following steps: 1) setting the charging connector of the electric vehicle to cooperate with the quick charging interface of the self-adjusting charging pile; 2) when the quick charging interface and the charging connector of the electric vehicle cannot cooperate, moving the electric vehicle to a position close to the self-adjusting charging pile and sending a signal to the charging interface; 3) after receiving the signal, the connector slides vertically along the vertical support column, and the charging interface follows the sliding block to slide along the slide groove; 4) when the charging interface follows the sliding block to slide along the slide groove, the wire on the winding device can be pulled out or retracted; 5) the charging interface and the charging connector of the electric vehicle cooperate with each other, and charging begins. However, in the prior art, only a few automated solutions mostly use pre-positioning tracks or simple mechanical push rods, which lack adaptability to vehicle parking position deviations and differences in charging port specifications and positions of different models, have poor robustness, and are difficult to apply on a large scale. Summary of the Invention

[0004] The purpose of this invention is to address the problems existing in the background technology by proposing a vision-guided adaptive alignment and positioning system and method for charging piles.

[0005] The technical solution of this invention: On the one hand, this application provides a vision-guided adaptive alignment and positioning method for charging piles, including: Vehicle information of multiple target vehicles of different models is collected, and the vehicle information is preprocessed to obtain preprocessed information; Create a positioning model; The preprocessed information is input into the localization model. The localization model determines the vehicle model based on the vehicle information and calculates the relative pose deviation between the charging plug and the charging port of the target vehicle by combining the 3D point cloud and the charging port position, so that the charging plug is inserted into the charging port, and the trained localization model is obtained. Real-time vehicle information is collected and input into the trained localization model to complete the charging of the real-time vehicles.

[0006] Preferably, the step of collecting vehicle information from multiple target vehicles of different models and preprocessing the vehicle information to obtain preprocessed information includes: Create a vehicle database; Set the acquisition parameters, collect vehicle information of multiple target vehicles of different models based on the acquisition parameters, and input all the collected data information into the vehicle database; the vehicle information includes vehicle model information and vehicle charging port information; Randomly select the vehicle information of a target vehicle from the vehicle database; Determine if there is any missing information in the vehicle information of the target vehicle; If there is missing information in the vehicle information of the target vehicle, the missing information is filled in based on the mean. The system returns the vehicle information of a target vehicle randomly selected from the vehicle database until all target vehicles have been selected, resulting in multiple preprocessed information items. The preprocessed information includes preprocessed model information and preprocessed charging port information.

[0007] Preferably, the step of inputting preprocessed information into the localization model, determining the vehicle model based on vehicle information through the localization model, and calculating the relative pose deviation between the charging plug and the charging port of the target vehicle by combining 3D point cloud and charging port position, so that the charging plug is inserted into the charging port, to obtain the trained localization model, includes: All preprocessed information is divided into training and test sets according to a random ratio; The training set is input into the localization model, and the monitoring model identifies the vehicle model of the target vehicle based on the vehicle model information, and obtains the location of the charging port of the target vehicle based on the vehicle model. By combining 3D point cloud and charging port position, the relative pose deviation between the charging plug and the charging port of the target vehicle is calculated, so that the charging plug is inserted into the charging port, and the trained localization model is obtained. Input the test set into the trained localization model to verify whether the localization model has been trained successfully.

[0008] Preferably, the step of inputting the training set into the localization model, identifying the vehicle model of the target vehicle based on the vehicle model information through the monitoring model, and obtaining the location of the charging port of the target vehicle based on the vehicle model includes: Preprocessed information of a target vehicle is randomly selected from the training set; The vehicle model of the target vehicle is identified by combining YOLO with pre-processed model information; Find the location of the standard charging port for the target vehicle based on its model number; Return the preprocessed information of a target vehicle randomly selected from the training set, until all target vehicles in the training set have been selected, and obtain the standard charging port location of each target vehicle.

[0009] Preferably, the step of combining 3D point cloud data and the charging port position to calculate the relative pose deviation between the charging plug and the charging port of the target vehicle, so that the charging plug is inserted into the charging port, to obtain the trained localization model, includes: Randomly select a target vehicle and obtain its preprocessed information and the location of its standard charging port; Calculate the actual location of the standard charging port of the target vehicle by combining the standard charging port location and the pre-processed charging port information; record the actual location of the standard charging port of the target vehicle as the standard location. Establish a unified coordinate system and include the position of the charging plug and the standard position within the unified coordinate system; Set a distance threshold; Obtain the position of the charging plug and the center coordinates of its standard position; Calculate the distance between the location of the charging plug and the center coordinates of the standard location, and determine whether the distance between the location of the charging plug and the center coordinates of the standard location is greater than or equal to the distance threshold. If the distance between the position of the charging plug and the center coordinates of the standard position is greater than or equal to the distance threshold, then move the charging plug and return to obtain the position of the charging plug and the center coordinates of the standard position until the distance between the position of the charging plug and the center coordinates of the standard position is less than the distance threshold. If the distance between the location of the charging plug and the center coordinates of the standard location is less than the distance threshold, then move the charging plug to the standard charging port location of the target vehicle.

[0010] Preferably, the step of moving the charging plug and returning to the point where the center coordinates of the charging plug and the standard position are greater than or equal to a distance threshold, until the distance between the center coordinates of the charging plug and the standard position is less than the distance threshold, includes: Obtain the x-coordinate and y-coordinate of the center of the charging plug, as well as the x-coordinate and y-coordinate of the center of the standard position; Calculate the difference between the horizontal coordinate of the center of the charging plug and the horizontal coordinate of the center of the standard position, and the difference between the vertical coordinate of the center of the charging plug and the vertical coordinate of the center of the standard position, respectively. Determine if the difference is greater than 0; If the difference is greater than 0, the charging plug will be moved along the positive direction of the coordinate axis corresponding to that coordinate. If the difference is less than 0, the charging plug will be moved in the negative direction of the coordinate axis corresponding to that coordinate.

[0011] Preferably, the step of collecting real-time vehicle information and inputting it into the trained localization model to complete the charging of the real-time vehicle includes: Collect real-time vehicle information; the real-time vehicle information includes real-time model information and real-time charging port information; Real-time vehicle information is input into the trained localization model, and the vehicle model and standard charging port location of the real-time vehicle are obtained through the trained localization model; the standard charging port of the real-time vehicle is recorded as the real-time charging port. Calculate the pose distance between the charging plug and the real-time charging port, and adjust the position of the charging plug so that it is inserted into the real-time charging port.

[0012] Preferably, the step of calculating the pose distance between the charging plug and the real-time charging port to adjust the position of the charging plug so that the charging plug is inserted into the real-time charging port includes: Obtain real-time location information of the charging plug and the charging port; The real-time distance between the charging plug and the charging port is calculated based on the location information. Determine whether the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold; If the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold, then insert the charging plug into the real-time charging port. If the real-time distance between the real-time charging plug and the real-time charging port is greater than or equal to the distance threshold, then adjust the position of the real-time charging plug until the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold.

[0013] On the other hand, this application also provides a vision-guided adaptive alignment and positioning system for charging piles, including a data acquisition component and a positioning component. The data acquisition component acquires vehicle information of the target vehicle and the position information of the charging plug. The positioning component executes the vision-guided adaptive alignment and positioning method for charging piles described above. All data information acquired by the data acquisition component is input into the positioning component. The positioning component adjusts the charging plug based on the position of the charging plug and the position of the charging port of the target vehicle, so that the charging plug can be automatically inserted into the charging port of the target vehicle.

[0014] Preferably, the acquisition component includes a vehicle acquisition module and a plug acquisition module. The vehicle acquisition module acquires vehicle information of the target vehicle, and the plug acquisition module acquires the location information of the charging plug.

[0015] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: By collecting vehicle information from multiple target vehicles of different models and preprocessing the information to obtain preprocessed information, a positioning model is created. The preprocessed information is then input into the positioning model, which determines the vehicle model based on the vehicle information. The relative pose deviation between the charging plug and the charging port of the target vehicle is calculated by combining 3D point cloud data and the charging port location, ensuring the charging plug is inserted into the charging port. This results in a trained positioning model. Finally, real-time vehicle information is collected and input into the trained positioning model to complete the charging of the real-time vehicle. This application introduces a scheme combining 3D point cloud technology and visual guidance to calculate accurate relative pose deviations and implements a closed-loop feedback movement strategy based on coordinate differences. This achieves sub-centimeter or even millimeter-level precise positioning of the charging plug, fundamentally avoiding charging failures, equipment damage, and safety accidents caused by misalignment, significantly improving the success rate of the charging process and the overall reliability of the system. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a vision-guided adaptive alignment and positioning method for charging piles proposed in this invention. Figure 2 This is a schematic diagram of the principle of a vision-guided adaptive alignment and positioning system for charging piles proposed in this invention. Figure descriptions: 100, data acquisition component; 101, vehicle data acquisition module; 102, plug data acquisition module; 200. Positioning component. Detailed Implementation

[0017] Example 1, as Figure 1 As shown, the present invention proposes a vision-guided adaptive alignment and positioning method for charging piles, comprising: S100 collects vehicle information from multiple target vehicles of different models and preprocesses the vehicle information to obtain preprocessed information. S200, Create a positioning model; S300 inputs preprocessed information into the localization model. The localization model determines the vehicle model based on vehicle information and calculates the relative pose deviation between the charging plug and the charging port of the target vehicle by combining 3D point cloud and charging port position, so that the charging plug is inserted into the charging port, and the trained localization model is obtained. The S400 collects real-time vehicle information and inputs it into the trained localization model to complete the charging of the real-time vehicle.

[0018] In this invention, vehicle information of multiple target vehicles of different models is collected and preprocessed to obtain preprocessed information. Then, a positioning model is created, and the preprocessed information is input into the positioning model. The positioning model determines the vehicle model based on the vehicle information and calculates the relative pose deviation between the charging plug and the charging port of the target vehicle by combining 3D point cloud and charging port position, so that the charging plug is inserted into the charging port. This results in a trained positioning model. Finally, real-time vehicle information of the real-time vehicle is collected and input into the trained positioning model to complete the charging of the real-time vehicle. This application introduces a scheme combining 3D point cloud technology and visual guidance to calculate accurate relative pose deviation and implements a closed-loop feedback movement strategy based on coordinate difference. This achieves sub-centimeter or even millimeter-level accurate positioning of the charging plug, fundamentally avoiding charging failure, equipment damage and safety accidents caused by misalignment, and significantly improving the success rate of the charging process and the overall reliability of the system.

[0019] In an optional embodiment, S100 includes: S110, Create vehicle database; S120, Set the acquisition parameters, collect vehicle information of multiple target vehicles of different models based on the acquisition parameters, and input all the collected data information into the vehicle database; the vehicle information includes vehicle model information and vehicle charging port information; S130, randomly select vehicle information of a target vehicle from the vehicle database; S140, Determine whether there is any missing information in the vehicle information of the target vehicle; S150, If there is missing information in the vehicle information of the target vehicle, the missing information is filled in based on the mean. S160, return the vehicle information of a target vehicle randomly selected from the vehicle database, until all target vehicles have been selected, and obtain multiple preprocessed information; the preprocessed information includes preprocessed model information and preprocessed charging port information.

[0020] It should be noted that the database uses an SQL or NoSQL structure and contains multiple fields such as vehicle model, brand, year, charging port type, and standard coordinates. The data acquisition parameters include, for each vehicle model, the requirement to acquire at least 10 sets of high-definition images and 3D point cloud data from different angles such as the front, side, and rear of the vehicle. For the charging port area, millimeter-level precision local scanning is required.

[0021] When preprocessing the data, the system checks whether the required fields of each record (such as the X, Y, and Z coordinates of the charging port center point) are empty to determine whether there is missing information in the vehicle information of the target vehicle. If the point cloud data file of a certain angle of a vehicle is damaged and the coordinates are missing, i.e., there is missing information, the system will not discard the data directly, but will fill in the missing information based on the mean, and finally obtain multiple high-quality preprocessed information, which constitute a reliable foundation for subsequent model training.

[0022] In an optional embodiment, S300 includes: S310, divide all preprocessed information into training and test sets according to a random ratio; S320 inputs the training set into the localization model, and the monitoring model identifies the vehicle model of the target vehicle based on the vehicle model information, and obtains the location of the charging port of the target vehicle based on the vehicle model. S330 combines 3D point cloud and charging port position to calculate the relative pose deviation between the charging plug and the charging port of the target vehicle, so that the charging plug is inserted into the charging port, and the trained localization model is obtained. S340: Input the test set into the trained localization model to verify whether the trained localization model has been successfully trained.

[0023] It should be noted that when dividing the training set and the test set, the proportion of the training set should be greater than that of the test set to ensure that there are enough training samples in the training set. All preprocessed information should be divided into the training set and the test set according to a random ratio (e.g., 80% training, 20% testing) to ensure the objectivity of model training and evaluation.

[0024] The model identifies the target vehicle's model based on vehicle model information by using a localization model. By learning from a large amount of labeled (vehicle model) images and point cloud data, a mapping relationship from visual features to specific vehicle models is established. After identifying the vehicle model, the model retrieves the target vehicle's charging port location (i.e., standard location) from the database based on the vehicle model.

[0025] Next, the core pose calculation stage begins. By combining 3D point cloud data and the charging port position, the relative pose deviation between the charging plug and the target vehicle's charging port is calculated. The localization model registers the real-time acquired vehicle 3D point cloud with an idealized, error-free vehicle model (e.g., using the ICP algorithm) to find the vehicle's actual posture. Then, combined with the known standard position of the charging port, the true coordinates of the charging port in the current actual posture are deduced. Comparing these true coordinates with the preset initial coordinates of the charging plug within the system, the precise pose deviation vector can be calculated. Based on this deviation vector, the model outputs control signals to drive the virtual plug to perform a "move-insert" operation in the simulation environment, resulting in the trained localization model.

[0026] By inputting the test set into the trained localization model, since the test set data has never been seen by the model during the training process, the system runs the test set in a simulation environment and statistically analyzes the model's vehicle recognition error rate, average localization error, and the success rate of the final "virtual insertion" on the test set.

[0027] In an optional embodiment, S320 includes: S321, randomly select the preprocessed information of a target vehicle from the training set; S322, using YOLO combined with preprocessed model information to identify the vehicle model of the target vehicle; S323, Obtain the location of the standard charging port for the target vehicle based on the vehicle model; S324, return the preprocessed information of a target vehicle randomly selected from the training set, until all target vehicles in the training set have been selected, and obtain the standard charging port location of each target vehicle. Specifically, the YOLO model is a real-time object detection model that predicts the position and category of an object simultaneously through a single forward propagation and is widely used in the field of computer vision. It should be noted that after the training set is input into the model, the localization model uses online learning or batch stochastic gradient descent to process samples one by one or in batches. Preprocessed information of a target vehicle is randomly selected from the training set; for example, data for a "2023 target vehicle Y" is selected.

[0028] Next, the vehicle model of the target vehicle is identified by combining YOLO with preprocessed model information. The external image of the vehicle is input into a pre-trained YOLOv7 or YOLOv8 model. This model, based on the COCO dataset, adds hundreds of additional car brand recognition categories. It can directly outline the vehicle from the image and give the recognition result of "2023 target vehicle Y". Then, based on the confirmed vehicle model, the location of the standard charging port of the target vehicle is obtained from the vehicle database or the built-in standard model.

[0029] Once all target vehicles in the training set have been selected for one or more rounds, the model has "seen" the standard locations of all vehicle types, obtained a mapping table of the standard charging port locations for each target vehicle, and internalized it into the model's weights.

[0030] In an optional embodiment, S330 includes: S331 randomly selects a target vehicle and obtains the preprocessing information of the target vehicle and the location of the standard charging port. S332, combine the standard charging port location with the pre-processed charging port information to calculate the actual location of the standard charging port of the target vehicle; record the actual location of the standard charging port of the target vehicle as the standard location; S333, Establish a unified coordinate system and place the position of the charging plug and the standard position into the unified coordinate system; S334, Set distance threshold; S335, obtain the position of the charging plug and the center coordinates of the standard position; S336, calculate the distance between the position of the charging plug and the center coordinates of the standard position, and determine whether the distance between the position of the charging plug and the center coordinates of the standard position is greater than or equal to the distance threshold. S337, If the distance between the position of the charging plug and the center coordinates of the standard position is greater than or equal to the distance threshold, then move the charging plug and return to obtain the position of the charging plug and the center coordinates of the standard position until the distance between the position of the charging plug and the center coordinates of the standard position is less than the distance threshold. S338, if the distance between the position of the charging plug and the center coordinate of the standard position is less than the distance threshold, then move the charging plug to the standard charging port position of the target vehicle.

[0031] It should be noted that in the computational stage of model training or practical application, a target vehicle is randomly selected (it can be a sample in the training set or a real-time vehicle), and the preprocessing information of the target vehicle and the location of the standard charging port are obtained.

[0032] Because actual vehicles may experience positional deviations, body tilt, or suspension changes when parked, directly using the standard position is inaccurate. Therefore, it is necessary to combine the standard charging port position with preprocessed charging port information (here referring to the noisy charging port observation position extracted from real-time 3D point clouds) to calculate the actual position of the target vehicle's standard charging port. This can be achieved through a sensor fusion or filtering algorithm. For example, using an Extended Kalman Filter (EKF), the standard position can be used as an ideal prior estimate, and the observation position extracted from the 3D point cloud can be used as a noisy measurement value. By fusing these values ​​through a filtering algorithm, a more accurate estimate that is closer to the true physical position can be obtained.

[0033] For ease of calculation, the system establishes a unified coordinate system. For example, an XYZ Cartesian coordinate system is established with the center of the charging pile base as the origin (0,0,0). The position of the charging plug and the standard position are both placed within this unified coordinate system. Simultaneously, a distance threshold is set, for example, 10mm, as the critical value for determining successful alignment.

[0034] The system uses sensors to acquire the real-time position of the charging plug and the center coordinates of the standard position (e.g., plug coordinates P_plug=(x_p, y_p, z_p), standard position coordinates P_target=(x_t, y_t, z_t)). It then calculates the distance between the charging plug's position and the center coordinates of the standard position, typically using Euclidean distance: Distance = sqrt((x_t - x_p)^2 + (y_t - y_p)^2 + (z_t - z_p)^2). Finally, it determines whether the distance between the charging plug's position and the center coordinates of the standard position is greater than or equal to a distance threshold.

[0035] If Distance >= 10mm, it indicates misalignment, and the charging plug is moved. After movement, the system will acquire the new plug coordinates in real time, recalculate the distance, and return the center coordinates of the charging plug's position and the standard position to begin a new round of judgment. This process is repeated iteratively to form a closed-loop control until Distance < 10mm.

[0036] Once the distance is less than 10mm, the system considers the plug's projection point on the horizontal plane to be sufficiently close to the target point. At this point, the charging plug is moved to the standard charging port position of the target vehicle. This step mainly refers to performing final fine-tuning of the attitude (such as rotation around the Z-axis and fine-tuning of the Z-axis height) to ensure that the plug can be accurately inserted into the charging port, completing the entire "insertion" action.

[0037] In an optional embodiment, S337 includes: S3371, Obtain the center x-coordinate and center y-coordinate of the charging plug, as well as the center x-coordinate and center y-coordinate of the standard position; S3372, calculate the difference between the center abscissa of the charging plug and the center abscissa of the standard position, and the difference between the center ordinate of the charging plug and the center ordinate of the standard position, respectively. S3373, determine whether the difference is greater than 0; S3374, if the difference is greater than 0, then move the charging plug along the positive direction of the coordinate axis corresponding to that coordinate. S3375, if the difference is less than 0, then move the charging plug along the negative direction of the coordinate axis corresponding to that coordinate.

[0038] It should be noted that when the system determines that the charging plug needs to be moved (i.e., Distance >= distance threshold), it executes the following fine-grained control strategy: First, obtain the center x-coordinate and center y-coordinate of the charging plug (x_p, y_p) and the center x-coordinate and center y-coordinate of the standard position (x_t, y_t). Then, calculate the difference between the center x-coordinate of the charging plug and the center x-coordinate of the standard position, dx = x_t - x_p, and the difference between the center y-coordinate of the charging plug and the center y-coordinate of the standard position, dy = y_t - y_p. Next, determine if the difference is greater than 0. For dx, if the difference dx>0, it means that the standard position is to the right of the current position of the plug, so the charging plug is moved along the positive X-axis (e.g., the X-axis motor of the robotic arm is controlled to rotate in the positive direction).

[0039] If the difference dx < 0, it means that the standard position is to the left of the current position of the plug, so the charging plug will be moved along the negative X-axis.

[0040] For dy, if the difference dy>0, it means that the standard position is in front of the current position of the plug, so the charging plug is moved along the positive Y-axis.

[0041] If the difference dy < 0, it means that the standard position is behind the current position of the plug, so the charging plug will be moved along the negative Y-axis.

[0042] The same judgment logic is used to control the height difference dz along the Z-axis. Through this step-by-step, axis-by-axis closed-loop feedback control, the charging plug can smoothly and accurately approach the target position in a manner similar to "PID control," avoiding the risk of collision or positioning overshoot caused by large-scale blind movement.

[0043] In an optional embodiment, S400 includes: S410, collect real-time vehicle information; the real-time vehicle information includes real-time model information and real-time charging port information; S420 inputs real-time vehicle information into the trained positioning model, and obtains the vehicle model and standard charging port location of the real-time vehicle through the trained positioning model; the standard charging port of the real-time vehicle is recorded as the real-time charging port. S430 calculates the positional distance between the charging plug and the real-time charging port to adjust the position of the charging plug so that it is inserted into the real-time charging port.

[0044] It should be noted that once the user has parked the car, the charging station begins operation, collecting real-time vehicle information. The dual-lens camera on top of the charging station and the LiDAR system surrounding the vehicle simultaneously activate, capturing images of the vehicle's exterior and a complete 3D point cloud. This real-time vehicle information includes the vehicle model and charging port information.

[0045] Real-time model information can be obtained in two ways: first, by reading the license plate through a license plate recognition (LPR) system and querying vehicle registration information online; second, by directly recognizing real-time images using the YOLO vision model. Real-time charging port information is obtained through semantic segmentation and feature extraction of the charging port area in the 3D point cloud.

[0046] Subsequently, real-time vehicle information is input into the trained localization model. Since the model has already been trained on massive amounts of data, its response speed is extremely fast (typically within a few hundred milliseconds). The trained localization model obtains the vehicle model and the location of the standard charging port. The model identifies this as a "NIO ET5" and retrieves the standard location of the charging port on the left rear fender from its knowledge base. The standard charging port of the real-time vehicle is recorded as the real-time charging port; here, "real-time" refers to the uniquely determined spatial coordinates of the charging port at the current moment, combined with the vehicle's actual parking posture.

[0047] Finally, the system calculates the pose distance between the charging plug and the real-time charging port (a comprehensive deviation including translation and rotation) to adjust the position of the charging plug. The positioning component decomposes the pose distance into control commands for each degree of freedom and sends them to the robotic arm controller, driving the robotic arm to move in coordination, continuously correcting the deviation, and ultimately inserting the charging plug into the real-time charging port, completing automatic alignment, and triggering the internal locking and power-on mechanism to begin charging the user.

[0048] In an optional embodiment, S430 includes: S431, obtain real-time location information of the charging plug and the charging port; S432, calculate the real-time distance between the real-time charging plug and the real-time charging port based on the location information; S433, determine whether the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold; S434, If the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold, then insert the charging plug into the real-time charging port. S435, if the real-time distance between the real-time charging plug and the real-time charging port is greater than or equal to the distance threshold, then adjust the position of the real-time charging plug until the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold.

[0049] It should be noted that the system obtains real-time location information for both the charging plug and the charging port. At this point, the plug is very close to the charging port.

[0050] The real-time distance between the charging plug and the charging port is calculated based on location information. This distance can be more precisely defined as the shortest straight-line distance between the end of the plug (the part closest to the charging port) and the center point of the charging port entrance.

[0051] It determines whether the real-time distance between the charging plug and the charging port is less than a distance threshold. This distance threshold is set very strictly, for example, to 2mm, to ensure gentle and safe insertion and prevent hard collisions from damaging the expensive charging plug or the vehicle's charging port.

[0052] If the real-time distance between the charging plug and the charging port is less than a distance threshold (e.g., <2mm), the system determines that it is in the optimal insertion window and inserts the charging plug into the charging port. This is usually a high-speed, short-stroke linear insertion action.

[0053] If the real-time distance between the charging plug and the charging port is greater than or equal to a distance threshold, the system will not insert the plug. Instead, it will continue to adjust the position of the charging plug (perhaps fine-tuning its attitude or performing final obstacle avoidance detection) until the real-time distance between the charging plug and the charging port is less than the distance threshold. This process ensures that physical contact only occurs under absolutely safe and highly accurate alignment conditions, greatly improving the system's reliability and lifespan.

[0054] like Figure 2As shown, this application also provides a visually guided adaptive alignment and positioning system for charging piles, including a data acquisition component 100 and a positioning component 200. The data acquisition component 100 acquires vehicle information of the target vehicle and the position information of the charging plug. The positioning component 200 executes the visually guided adaptive alignment and positioning method for charging piles described above. All data information acquired by the data acquisition component 100 is input into the positioning component 200. The positioning component 200 adjusts the charging plug based on the position of the charging plug and the position of the charging port of the target vehicle, so that the charging plug can be automatically inserted into the charging port of the target vehicle.

[0055] It should be noted that the acquisition component 100 of this application consists of 2-4 high-resolution industrial cameras mounted on a rotatable gimbal on the top of the charging pile, one or more 16-line / 32-line LiDARs and an inertial measurement unit. The plug acquisition module 102 is set on the base and end of the charging plug and integrates multiple miniature photoelectric encoders or magnetic scales for real-time and high-precision acquisition of the position information of the charging plug.

[0056] The positioning component 200 is a ruggedized industrial control computer or a high-performance embedded GPU platform. The acquisition component 100 inputs all the acquired data information into the positioning component 200. After receiving the data, the positioning component 200 performs real-time analysis and calculation, and then sends high-frequency control commands to the robotic arm of the charging pile. By combining the position of the charging plug and the position of the charging port of the target vehicle, the positioning component 200 adjusts the charging plug and controls the robotic arm to perform complex spatial movements, ultimately realizing the function of automatically inserting the charging plug into the charging port of the target vehicle.

[0057] In an optional embodiment, the acquisition component 100 includes a vehicle acquisition module 101 and a plug acquisition module 102. The vehicle acquisition module 101 acquires vehicle information of the target vehicle, and the plug acquisition module 102 acquires the location information of the charging plug.

[0058] It should be noted that the vehicle acquisition module 101 is a vision and perception module, which is used to capture images, control the LiDAR to perform 360-degree scanning, and perform preliminary stitching, noise reduction and formatting of the raw data stream. The vehicle acquisition module 101 collects vehicle information of the target vehicle. The vehicle information is specific optical image data and three-dimensional point cloud data, which are the raw inputs for subsequent vehicle type recognition and pose calculation.

[0059] The core of the plug acquisition module 102 is the plug status monitoring module. It reads the output signal of the encoder or magnetic ruler and converts it into a digital signal representing the real-time spatial pose of the charging plug. Therefore, the position information of the charging plug acquired by the plug acquisition module 102 is the precise pose data of the end effector of the robotic arm, which is one of the benchmarks for closed-loop position control and pose deviation calculation.

[0060] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A visually guided adaptive alignment and positioning method for charging piles, characterized in that, include: Vehicle information of multiple target vehicles of different models is collected and preprocessed to obtain preprocessed information. Create a positioning model; The preprocessed information is input into the localization model. The localization model determines the vehicle model based on the vehicle information and calculates the relative pose deviation between the charging plug and the charging port of the target vehicle by combining the 3D point cloud and the charging port position, so that the charging plug is inserted into the charging port, and the trained localization model is obtained. Real-time vehicle information is collected and input into the trained localization model to complete the charging of the real-time vehicles.

2. The visually guided adaptive alignment and positioning method for charging piles according to claim 1, characterized in that, The process involves collecting vehicle information from multiple target vehicles of different models and preprocessing the vehicle information to obtain preprocessed information, including: Create a vehicle database; Set the acquisition parameters, collect vehicle information of multiple target vehicles of different models based on the acquisition parameters, and input all the collected data information into the vehicle database; the vehicle information includes vehicle model information and vehicle charging port information; Randomly select the vehicle information of a target vehicle from the vehicle database; Determine if there is any missing information in the vehicle information of the target vehicle; If there is missing information in the vehicle information of the target vehicle, the missing information is filled in based on the mean. The system returns the vehicle information of a target vehicle randomly selected from the vehicle database until all target vehicles have been selected, resulting in multiple preprocessed information items. The preprocessed information includes preprocessed model information and preprocessed charging port information.

3. The visually guided adaptive alignment and positioning method for charging piles according to claim 2, characterized in that, The process involves inputting preprocessed information into a localization model, determining the vehicle model based on vehicle information, and calculating the relative pose deviation between the charging plug and the charging port of the target vehicle using 3D point cloud data and the charging port location. This ensures the charging plug is inserted into the charging port, resulting in a trained localization model. All preprocessed information is divided into training and test sets according to a random ratio; The training set is input into the localization model, and the monitoring model identifies the vehicle model of the target vehicle based on the vehicle model information, and obtains the location of the charging port of the target vehicle based on the vehicle model. By combining 3D point cloud and charging port position, the relative pose deviation between the charging plug and the charging port of the target vehicle is calculated, so that the charging plug is inserted into the charging port, and the trained localization model is obtained. Input the test set into the trained localization model to verify whether the localization model has been trained successfully.

4. The visually guided adaptive alignment and positioning method for charging piles according to claim 3, characterized in that, The step of inputting the training set into the localization model, identifying the target vehicle's model based on vehicle model information through the monitoring model, and obtaining the target vehicle's charging port location based on the vehicle model includes: Preprocessed information of a target vehicle is randomly selected from the training set; The vehicle model of the target vehicle is identified by combining YOLO with pre-processed model information; Find the location of the standard charging port for the target vehicle based on its model number; Return the preprocessed information of a target vehicle randomly selected from the training set, until all target vehicles in the training set have been selected, and obtain the standard charging port location of each target vehicle.

5. The visually guided adaptive alignment and positioning method for charging piles according to claim 4, characterized in that, The process of combining 3D point cloud data and the charging port position to calculate the relative pose deviation between the charging plug and the charging port of the target vehicle, ensuring the charging plug is inserted into the charging port, and obtaining the trained localization model includes: Randomly select a target vehicle and obtain its preprocessed information and the location of its standard charging port; Calculate the actual location of the standard charging port of the target vehicle by combining the standard charging port location and the pre-processed charging port information; record the actual location of the standard charging port of the target vehicle as the standard location. Establish a unified coordinate system and include the position of the charging plug and the standard position within the unified coordinate system; Set a distance threshold; Obtain the position of the charging plug and the center coordinates of its standard position; Calculate the distance between the location of the charging plug and the center coordinates of the standard location, and determine whether the distance between the location of the charging plug and the center coordinates of the standard location is greater than or equal to the distance threshold. If the distance between the position of the charging plug and the center coordinates of the standard position is greater than or equal to the distance threshold, then move the charging plug and return to obtain the position of the charging plug and the center coordinates of the standard position until the distance between the position of the charging plug and the center coordinates of the standard position is less than the distance threshold. If the distance between the location of the charging plug and the center coordinates of the standard location is less than the distance threshold, then move the charging plug to the standard charging port location of the target vehicle.

6. The visually guided adaptive alignment and positioning method for charging piles according to claim 5, characterized in that, If the distance between the position of the charging plug and the center coordinates of the standard position is greater than or equal to a distance threshold, then the charging plug is moved, and the process of obtaining the position of the charging plug and the center coordinates of the standard position is repeated until the distance between the position of the charging plug and the center coordinates of the standard position is less than the distance threshold, including: Obtain the x-coordinate and y-coordinate of the center of the charging plug, as well as the x-coordinate and y-coordinate of the center of the standard position; Calculate the difference between the horizontal coordinate of the center of the charging plug and the horizontal coordinate of the center of the standard position, and the difference between the vertical coordinate of the center of the charging plug and the vertical coordinate of the center of the standard position, respectively. Determine if the difference is greater than 0; If the difference is greater than 0, the charging plug will be moved along the positive direction of the coordinate axis corresponding to that coordinate. If the difference is less than 0, the charging plug will be moved in the negative direction of the coordinate axis corresponding to that coordinate.

7. The visually guided adaptive alignment and positioning method for charging piles according to claim 6, characterized in that, The process of collecting real-time vehicle information and inputting it into a trained localization model to charge the real-time vehicles includes: Collect real-time vehicle information; the real-time vehicle information includes real-time model information and real-time charging port information; Real-time vehicle information is input into the trained localization model, and the vehicle model and standard charging port location of the real-time vehicle are obtained through the trained localization model; the standard charging port of the real-time vehicle is recorded as the real-time charging port. Calculate the pose distance between the charging plug and the real-time charging port, and adjust the position of the charging plug so that it is inserted into the real-time charging port.

8. The visually guided adaptive alignment and positioning method for charging piles according to claim 7, characterized in that, The calculation of the pose distance between the charging plug and the real-time charging port, in order to adjust the position of the charging plug so that the charging plug is inserted into the real-time charging port, includes: Obtain real-time location information of the charging plug and the charging port; The real-time distance between the charging plug and the charging port is calculated based on the location information. Determine whether the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold; If the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold, then insert the charging plug into the real-time charging port. If the real-time distance between the real-time charging plug and the real-time charging port is greater than or equal to the distance threshold, then adjust the position of the real-time charging plug until the real-time distance between the real-time charging plug and the real-time charging port is less than the distance threshold.

9. A vision-guided adaptive alignment and positioning system for charging piles, characterized in that, include: Data acquisition components; The acquisition component collects vehicle information of the target vehicle and the location information of the charging plug. Positioning components; The positioning component executes the visually guided adaptive alignment and positioning method for charging piles as described in any one of claims 1 to 8. All data information collected by the acquisition component is input into the positioning component. The positioning component adjusts the charging plug in conjunction with the position of the charging plug and the position of the charging port of the target vehicle, so that the charging plug can be automatically inserted into the charging port of the target vehicle.

10. A vision-guided adaptive alignment and positioning storage device for charging piles, characterized in that, The data acquisition components include a vehicle acquisition module and a plug acquisition module. The vehicle acquisition module acquires vehicle information of the target vehicle, and the plug acquisition module acquires the location information of the charging plug.

Images