A charging gun plug-pull fatigue compensation and pose self-adaptive correction method

Abstract

The application discloses a charging gun plug-in fatigue compensation and pose self-adaptive correction method and belongs to the technical field of robot control. The prior art lacks an online identification and compensation mechanism for accumulated errors of a mechanical arm joint in long-term plug-in operation of a charging robot. The application realizes real-time acquisition of force sensation data in the plug-in process through a force sensor; a mechanical arm joint error accumulation model is constructed, and pose deviation caused by joint clearance and elastic deformation is identified online based on historical plug-in data; joint fatigue compensation is calculated according to the current plug-in frequency and force sensation feedback; the compensation is superimposed on the target pose to generate a corrected plug-in trajectory; and the pose correction amount is dynamically adjusted based on real-time force sensation feedback during the plug-in process. The application realizes online identification and self-adaptive compensation of pose errors in long-term plug-in operation of the charging robot, and significantly prolongs the precision maintenance period of the mechanical arm and the service life of the charging gun.

Description

Technical Field

[0003] This invention belongs to the field of robot control technology, specifically relating to a charging gun plugging and unplugging control method for a new energy vehicle charging robot, and more particularly to a technical solution for online identification and adaptive compensation of cumulative errors in the joints of a robotic arm during long-term repetitive plugging and unplugging operations. Background Technology

[0005] With the rapid development of automated charging robots, these robots need to perform a large number of plugging and unplugging operations in charging station scenarios. Taking BYD's flash charging station as an example, a single charging robot may need to perform dozens or even hundreds of plugging and unplugging operations per day. In long-term repetitive plugging and unplugging operations, the joint clearance, elastic deformation, and wear of the robotic arm will cause the positional accuracy of the end effector to gradually decrease, directly affecting the success rate and lifespan of the docking of the charging gun with the vehicle's charging port.

[0006] The prior art related to this application, as determined by a search, is as follows:

[0007] I. Force-based insertion and removal control technology

[0008] Prior art document 1: PCT patent application No. WO2023116667A1 (filed on December 29, 2022) discloses a method for controlling the charging of a robotic arm. Its technical solution is as follows: a force sensor collects force information of the contact surface between the charging plug and the charging port; pose adjustment parameters are determined based on the force information; the pose of the robotic arm body is corrected from a first pose to a second pose; wherein the second pose is the pose when the charging head is aligned with the charging port. This technology utilizes force feedback for pose correction during a single insertion / removal process, but it does not address the identification and compensation of cumulative joint errors during long-term repetitive insertion / removal operations.

[0009] Prior art document 2: Chinese patent application CN119682571A (Voyah Automobile Technology Co., Ltd., application date December 24, 2024) discloses a vehicle charging control method. Its technical solution is as follows: determining the force state of the charging gun during the insertion and removal of the charging port; determining whether there is a change in force or torque greater than a preset threshold based on the force state; if so, adjusting the posture of the charging gun based on the change value and the preset initial force value. This technology is also a force feedback control during a single insertion and removal process and does not involve fatigue compensation for long-term operation.

[0010] Reference document 3: Chinese patent application CN106335057A (Southeast University, application date September 27, 2016) discloses a method for full-space compliant socket control of an assembly robot based on real-time force control. Its technical solution is as follows: a real-time force control system is built using a six-dimensional wrist force sensor; a gravity compensation method based on standard position is used to eliminate the influence of gravity; and force / position hybrid control is performed on the four stages of the socket task: approach, socket finding, insertion, and insertion completion. This technology is for compliant socket control in general assembly scenarios, but it does not address the specific scenarios of charging robots, nor does it address the compensation for accumulated joint errors in long-term repetitive operations.

[0011] II. Pose Correction Technology Based on Visual and Force Fusion

[0012] Reference document 4: Chinese patent application CN121608159A (Chenzhi Automotive Technology Group Co., Ltd., application date January 21, 2026) discloses a charging robot plug-in method. Its technical solution is as follows: when the end effector of the robotic arm reaches a preset position in front of the target charging port, multi-stage approximation control comprising at least two positioning stages is executed; in each positioning stage, the current position deviation between the end effector of the robotic arm and the target charging port is acquired; if the deviation is less than a threshold, the next stage is entered; otherwise, posture adjustment is performed based on an adjustment coefficient. This technology involves posture deviation correction in multi-stage approximation control, but does not involve modeling and compensation for cumulative joint errors during long-term repeated plugging and unplugging.

[0013] Prior art document 5: Chinese patent application CN120792570A (filed on June 27, 2025) discloses a charging pile cable plugging and unplugging system based on machine vision. Its technical solution involves: capturing cable surface strain distribution data using multi-view vision sensors; establishing an error propagation matrix including spatiotemporal correlation characteristics; and generating the pose compensation amount for the end effector. This technology involves vision-based pose compensation, but the compensation object is the error caused by cable deformation, rather than the long-term cumulative error of the robotic arm joints.

[0014] III. Robotic Arm Precision Compensation Technology

[0015] Reference document 6: A Chinese patent (publication date March 29, 2024) discloses a technical solution for a method of vision-guided precision operation of a robotic arm. This solution involves obtaining a compensation matrix that can compensate for errors in the robotic arm's parameters, effectively compensating for the negative effects caused by low-precision parameters. While this technology addresses the compensation of robotic arm parameter errors, the compensation matrix is ​​a static calibration result and does not involve online identification and dynamic updating based on historical insertion / removal data.

[0016] Prior art document 7: Chinese patent application CN118942080A (Suzhou Dolphin Star Intelligent Technology Co., Ltd.) discloses an autonomous charging method for unmanned forklifts that can accurately visually identify the pose of the charging port, utilizing semantic segmentation for charging port feature extraction and pose estimation. This technology involves the visual localization of the charging port but does not address compensation for joint errors in the robotic arm.

[0017] IV. Joint Wear and Fatigue Related Technologies

[0018] Reference document 8: A Chinese patent (China Construction Bank, publication date April 29, 2026) concerning "a method for uniform wear motion control of a redundant robotic arm driven by joint acceleration" discloses a technical solution that constructs a parametric representation of the uniform wear of each joint in a redundant robotic arm. This technology involves uniform wear control of robotic arm joints, but its goal is "uniform wear" (to balance the wear of each joint through motion planning), rather than online identification and compensation for pose deviations caused by existing joint wear.

[0019] Overall shortcomings of existing technologies

[0020] Based on the aforementioned comparative documents, the prior art shares the following common defects:

[0021] 1. Lack of an online identification mechanism for cumulative joint errors during long-term repetitive insertion and removal operations. Existing technologies (comparative documents 1, 2, and 3) all rely on force feedback control during a single insertion and removal process, without considering the decline in pose accuracy of robotic arm joints due to increased clearance, elastic deformation, and wear after long-term use.

[0022] 2. Lack of a joint error accumulation model based on historical insertion and removal data. Although the existing technology (comparative document 6) involves compensation for robotic arm parameter errors, it is based on static calibration results and does not establish a dynamic update model based on historical data.

[0023] 3. There is a lack of a correction mechanism that combines joint fatigue compensation with real-time force feedback. In the prior art, force feedback control (reference documents 1, 2, 3) and accuracy compensation (reference documents 6, 8) are usually used as independent technologies, and no scheme that organically combines the two has been found.

[0024] 4. Existing technologies are not specifically optimized for the high-frequency plugging and unplugging operations of charging robots. Charging robots need to perform dozens or even hundreds of plugging and unplugging operations every day, and the rate of joint fatigue accumulation is much higher than that of general industrial robot applications. Summary of the Invention

[0026] I. Technical Problems to be Solved

[0027] In view of the above-mentioned defects of the prior art, the present invention provides a charging gun plugging and unplugging fatigue compensation and posture adaptive correction method, which aims to solve the following technical problems: (1) how to identify the cumulative error of the robotic arm joint in long-term repetitive plugging and unplugging operations online; (2) how to establish a joint error accumulation model based on historical plugging and unplugging data; (3) how to combine joint fatigue compensation with real-time force feedback to achieve adaptive posture correction; (4) how to optimize for high-frequency plugging and unplugging scenarios of charging robots.

[0028] II. Technical Solution

[0029] This invention provides a method for fatigue compensation and posture adaptive correction of charging gun insertion and removal, comprising the following steps:

[0030] S1: Real-time force data acquisition. Force data during the insertion and removal process is acquired in real time through force sensors installed at the end of the charging robot's robotic arm or on the charging gun.

[0031] The force sensor includes a six-dimensional force sensor or a one-dimensional / multi-dimensional force sensor, which is installed between the end flange of the robotic arm and the charging gun clamping mechanism, or directly on the charging gun body. The force data includes: force in the insertion / removal direction (Z-axis), force perpendicular to the insertion / removal direction (X-axis, Y-axis), and torque about each axis.

[0032] S2: Joint Error Accumulation Model Construction and Online Identification. A joint error accumulation model for the robotic arm is constructed, and pose deviations caused by joint clearance and elastic deformation are identified online based on historical insertion and removal data.

[0033] The robotic arm joint error accumulation model includes the following three sub-models:

[0034] (1) Joint clearance model – based on the positional deviation of joints under no-load and loaded conditions. Each rotary or locating joint of the robotic arm has a certain mechanical backlash, which gradually increases after long-term use. The clearance model describes the relationship between the deviation between the joint angle command value and the actual angle value and the joint load state.

[0035] (2) Joint Elastic Deformation Model – Based on the relationship between joint force and deformation. During insertion and extraction, the robotic arm bears a large force and torque, and the transmission components of each joint (such as reducers, harmonic reducers, etc.) will undergo elastic deformation, resulting in end-effector pose deviation. The elastic deformation model describes the relationship between joint force and deformation.

[0036] (3) Joint wear and degradation model – based on the relationship between the number of insertions and removals and the decline in joint accuracy. As the number of insertions and removals increases, the transmission components of the joint will wear, leading to a gradual decrease in accuracy. The wear and degradation model describes the relationship between the cumulative number of insertions and removals and the decline in joint accuracy.

[0037] The specific method for online identification is as follows: After each insertion / removal operation, the deviation between the actual position and the theoretical target position of the robotic arm's end effector is recorded; based on the deviation data, the parameters of the joint error accumulation model are updated using either Recursive Least Squares (RLS) or Kalman Filter. By continuously accumulating historical data, the model parameters gradually converge to the true values, achieving continuous tracking of the joint state.

[0038] S3: Calculation of joint fatigue compensation. Calculate the joint fatigue compensation based on the current cumulative number of insertions and removals and the force sensation data.

[0039] Specifically, the calculation process consists of two steps:

[0040] (1) Predicted pose deviation query - Based on the current cumulative number of insertions and removals, query the joint error accumulation model (especially the wear degradation model) to obtain the predicted pose deviation of each joint under the current wear state.

[0041] (2) Force data correction - Based on the force data collected during the current insertion and extraction process, the force state is determined and the predicted pose deviation is corrected. For example, when the resistance in the insertion and extraction direction is too large, it indicates that there may be additional elastic deformation or jamming in the joint, and the compensation amount in the corresponding direction needs to be increased; when the lateral force is too large, it indicates that there may be lateral deviation at the end, and the compensation amount perpendicular to the insertion and extraction direction needs to be adjusted.

[0042] The formula for calculating the joint fatigue compensation amount is as follows:

[0043]

[0044] in, This represents the joint fatigue compensation amount, where N is the current cumulative number of insertions and removals, and F is the current force perception data. These are the current parameters of the joint error accumulation model.

[0045] S4: Generating the corrected insertion / removal trajectory. The joint fatigue compensation amount is superimposed on the target pose to generate the corrected insertion / removal trajectory.

[0046] Specifically: the joint fatigue compensation amount is decomposed into the angle compensation value of each joint; the angle compensation value of each joint is superimposed on the target angle of each joint to generate the corrected motion command of each joint.

[0047] The corrected target pose is:

[0048]

[0049] in, The original target pose, For the corrected target pose

[0050] S5: Dynamic adjustment of pose correction based on force feedback. During insertion and removal, the pose correction is dynamically adjusted based on real-time force feedback until insertion and removal are completed.

[0051] Specifically:

[0052] (1) When the resistance detected in the insertion / removal direction (Z-axis) exceeds the preset threshold, the pose correction amount in the corresponding direction is increased so that the charging gun can continue to be inserted at a more suitable angle and position;

[0053] (2) When a lateral (X-axis or Y-axis) force is detected to exceed a preset threshold, adjust the pose correction amount perpendicular to the insertion / removal direction to eliminate lateral deviation;

[0054] (3) When the torque around each axis is detected to exceed the preset threshold, adjust the attitude angle correction amount to eliminate the attitude deviation.

[0055] Preferably, the method further includes: when the cumulative force data during a single insertion / removal process exceeds a preset safety threshold, triggering a safety protection mechanism to stop the insertion / removal action and issue an alarm signal to prevent damage to the charging gun or the vehicle charging port.

[0056] III. Beneficial Effects

[0057] 1. A joint error accumulation model for long-term repetitive insertion and removal operations was established. Compared with the existing technology that only focuses on the force feedback control of a single insertion and removal operation (comparative documents 1, 2, and 3), this invention establishes a complete joint error accumulation model that includes a gap model, an elastic deformation model, and a wear degradation model, thereby achieving comprehensive tracking of the joint state.

[0058] 2. Online parameter identification based on historical insertion and removal data was achieved. Compared with the static calibration compensation scheme in the prior art (reference document 6), this invention adopts the recursive least squares method or Kalman filtering method to continuously update the model parameters based on the deviation data after each insertion and removal, so that the compensation amount can be adaptively adjusted according to the changes in joint state.

[0059] 3. Integrating joint fatigue compensation with real-time force feedback. Compared to the independent application of force feedback control (reference documents 1, 2, 3) and accuracy compensation (reference documents 6, 8) in existing technologies, this invention uses fatigue compensation based on historical data as a "feedforward" correction and real-time force feedback as a "feedback" correction. The two work together to achieve more precise pose control.

[0060] 4. Specifically designed for high-frequency plugging and unplugging scenarios of charging robots. The wear and degradation model of this invention is specifically designed for the high-frequency operation scenario of charging robots, which involves dozens or even hundreds of plugging and unplugging operations per day. This effectively extends the precision maintenance cycle of the robotic arm and the service life of the charging gun. Attached Figure Description

[0062] Figure 1 This is an overall flowchart of the charging gun plug-in fatigue compensation and posture adaptive correction method of the present invention.

[0063] Figure 2 This is a flowchart illustrating the steps of constructing and online identifying the joint error accumulation model of the robotic arm according to the present invention.

[0064] Figure 3 This is a flowchart illustrating the steps for calculating joint fatigue compensation in this invention.

[0065] Figure 4 This is a flowchart illustrating the dynamic adjustment steps of the pose correction amount based on force feedback according to the present invention. Detailed Implementation

[0067] Example 1: Force Feedback Control in Initial Insertion and Removal Operation

[0068] During the first plugging and unplugging operation after leaving the factory, the robotic arm joints have not yet shown obvious wear and clearance. At this time, the parameters of the joint error accumulation model are the initial calibration values ​​(based on the factory calibration data).

[0069] The system performs the plug-in / plug-out operation as follows:

[0070] The robotic arm carrying the charging gun moves towards the vehicle's charging port. When it approaches the charging port (approximately 5 centimeters away), the system switches the motion mode from position control to a force-position hybrid control mode.

[0071] The six-dimensional force sensor, located between the end flange of the robotic arm and the charging gun clamping mechanism, begins to collect force data in real time. The acquisition frequency is 1000Hz, and the data includes: Fx (force in the X-axis direction), Fy (force in the Y-axis direction), Fz (force in the Z-axis direction, i.e., the insertion and extraction direction), Mx, My, and Mz (torques around each axis).

[0072] During the insertion phase, the system monitors the resistance value in the Fz direction. When Fz rises from nearly 0N to approximately 10N, the system determines that the charging gun has begun to enter the charging port. At this point, the system reduces the insertion speed from 50mm / s to 20mm / s to avoid impact.

[0073] The system continuously monitors the lateral forces in the Fx and Fy directions. When Fx exceeds 5N, it determines that the charging gun has a deviation in the X direction, and the system automatically adjusts the pose correction amount by +0.5mm in the X direction; when Fy exceeds 5N, it adjusts the pose correction amount by +0.5mm in the Y direction.

[0074] When the resistance in the Fz direction reaches 50N (preset force control threshold), the system determines that the charging gun has been properly inserted. The system stops the feeding action, maintains the current position, and notifies the charging pile to start charging.

[0075] During the unplugging phase, the system pulls the charging gun backward at a speed of 20mm / s, monitoring the resistance in the Fz direction in real time. When the resistance approaches 0, it is determined that the charging gun has been completely detached from the charging port, completing the unplugging action.

[0076] After this insertion and removal operation is completed, the system records the deviation between the actual position of the end effector and the theoretical target position (the deviation in this case is +0.1mm, which is within the factory accuracy range). This data will be used as historical data for subsequent online updates of the model.

[0077] Example 2: Fatigue compensation after long-term insertion and removal

[0078] The charging robot has performed 10,000 plug-in / plug-out operations. With the increase in the number of plug-in / plug-out operations, the harmonic reducers of each joint of the robotic arm have experienced varying degrees of wear, resulting in increased joint clearances. This has caused the positioning accuracy of the end effector to decrease from ±0.05mm at the factory to ±0.25mm. This decrease in accuracy will affect the success rate of the charging gun docking with the vehicle's charging port.

[0079] After each insertion / removal operation, the system records the deviation between the actual position of the end effector and the theoretical target position. Based on this historical deviation data of 10,000 times, the system uses recursive least squares (RLS) to update the parameters of the joint error accumulation model online.

[0080] When the cumulative number of insertions and removals reaches 10,000, the system queries the joint wear and degradation model to obtain the predicted pose deviations of each joint. The model prediction results show that joint 1 (base rotation joint) has an angle deviation of approximately 0.08°, joint 2 (shoulder joint) has an angle deviation of approximately 0.12°, and joint 3 (elbow joint) has an angle deviation of approximately 0.06°. These deviations, when superimposed at the end effector, cause the end effector to shift by approximately 0.15mm in the X direction and approximately 0.08mm in the Y direction.

[0081] The system converts the predicted pose deviation into joint fatigue compensation amounts. The specific calculation process is as follows: Joint 1 compensation +0.08°, Joint 2 compensation +0.12°, Joint 3 compensation +0.06°. The direction of these compensation amounts is opposite to the direction of the deviation, used to offset pose errors caused by joint wear.

[0082] The system superimposes the compensation values ​​of each joint onto the target angle of each joint. For example, if the original target angles are 130.00°, 245.00°, and 360.00°, the corrected target angles are 130.08°, 245.12°, and 360.06°.

[0083] During insertion and extraction operations, the system dynamically adjusts the pose correction amount based on fatigue compensation and real-time force feedback. When the system detects that the resistance in the insertion / extraction direction reaches 45N (close to the 50N threshold) at an insertion depth of 5mm, which is higher than the 30N during the first insertion / extraction, the system determines that the joint elastic deformation or jamming is causing additional resistance, automatically reduces the insertion speed to 10mm / s, and increases the pose correction amount in the insertion / extraction direction by +0.2mm.

[0084] In the compensated plugging and unplugging operation, the actual positioning accuracy of the end effector was restored to about ±0.08mm, and the docking success rate of the charging gun and the vehicle charging port reached more than 99.5%.

[0085] After insertion and removal are completed, the system records the deviation data and continues to update the joint error accumulation model. The model parameters converge continuously as data accumulates, enabling continuous tracking of the joint state.

[0086] Example 3: Safety Protection When Force Data is Abnormal

[0087] When the Furman charging robot was performing a charging gun insertion operation, the six-dimensional force sensor detected that the resistance in the Fz direction reached 80N when the charging gun entered the charging port about 3mm, which far exceeded the preset safety threshold (60N).

[0088] The system determines that the current plugging / unplugging status is abnormal—this may be due to a serious misalignment between the charging gun and the charging port, foreign objects in the charging port, or damage to the vehicle's charging port.

[0089] The system immediately triggers a safety protection mechanism: it stops all movement of the robotic arm and holds the charging gun in its current position; it alerts on-site personnel via voice announcement "Charging plugging / unplugging abnormal, safely stopped"; it sends an alarm message (including a snapshot of force data and robot pose data at the time of the abnormality) to maintenance personnel via the network; and it awaits manual intervention.

[0090] The safety protection mechanism ensures that the charging gun or vehicle charging port will not be damaged in abnormal situations, thus protecting the safety of the equipment and the vehicle.

[0091] Example 4: Initialization and Convergence of Model Parameters

[0092] Before the first insertion and removal operation, the system initializes the joint error accumulation model. The initial parameters are set based on the robot arm's factory calibration data: the gaps between each joint are set to 0, the elastic deformation coefficients of each joint are set based on the stiffness data provided by the manufacturer, and the initial accuracy decay rate of the wear degradation model is set to 0 (assuming no wear at the time of manufacture).

[0093] As the number of insertions and removals increases, the system uses the pose deviation data after each operation to update the model parameters online. When the number of insertions and removals reaches 500, the model parameters begin to converge, and the root mean square error (RMSE) between the predicted pose deviation and the actual deviation decreases to below 0.05 mm. When the number of insertions and removals reaches 5000, the model parameters converge further, and the RMS error decreases to below 0.02 mm.

[0094] Through online identification, the model can accurately track the wear and degradation process of joints, providing a reliable basis for calculating fatigue compensation.

Claims

1. A method for fatigue compensation and adaptive posture correction during charging gun insertion and removal, characterized in that, Includes the following steps: S1: Force sensor installed at the end of the charging robot arm or on the charging gun collects force data in real time during the insertion and removal process; S2: Construct a cumulative joint error model for the robotic arm and identify pose deviations caused by joint clearance and elastic deformation online based on historical insertion and removal data; S3: Calculate the joint fatigue compensation amount based on the current cumulative number of insertions and removals and the force sensation data; S4: The joint fatigue compensation amount is superimposed on the target pose to generate the corrected insertion / removal trajectory; S5: During the insertion and removal process, the pose correction amount is dynamically adjusted based on real-time force feedback until the insertion and removal are completed.

2. The method according to claim 1, characterized in that, The robotic arm joint error accumulation model includes: a gap model for each joint, an elastic deformation model for each joint, and a wear degradation model for each joint; the gap model is established based on the positional deviation of the joint under no-load and loaded conditions, the elastic deformation model is established based on the relationship between the force and deformation of the joint, and the wear degradation model is established based on the relationship between the number of insertions and removals and the decrease in joint accuracy.

3. The method according to claim 1, characterized in that, The online identification includes: after each insertion and removal operation, recording the deviation between the actual position of the robotic arm end effector and the theoretical target position; based on the deviation data, updating the parameters of the joint error accumulation model using recursive least squares or Kalman filtering.

4. The method according to claim 1, characterized in that, The calculation of joint fatigue compensation in step S3 includes: querying the joint error accumulation model based on the current cumulative number of insertions and removals to obtain the predicted pose deviation of each joint; judging the force state during the current insertion and removal process based on the force data, correcting the predicted pose deviation, and obtaining the joint fatigue compensation amount.

5. The method according to claim 1, characterized in that, The step S4 of generating the corrected insertion and removal trajectory includes: decomposing the joint fatigue compensation amount into the angle compensation value of each joint; superimposing the angle compensation value of each joint onto the target angle of each joint to generate the corrected motion command of each joint.

6. The method according to claim 1, characterized in that, The dynamic adjustment of pose correction based on real-time force feedback in step S5 includes: increasing the pose correction in the corresponding direction when the resistance in the insertion / removal direction exceeds a preset threshold; adjusting the pose correction perpendicular to the insertion / removal direction when the lateral force exceeds a preset threshold; and adjusting the attitude angle correction when the torque exceeds a preset threshold.

7. The method according to claim 1, characterized in that, The method further includes: when the cumulative force data during a single insertion / removal process exceeds a preset safety threshold, a safety protection mechanism is triggered to stop the insertion / removal action and issue an alarm signal.

8. The method according to claim 1, characterized in that, The method is applied to a new energy vehicle charging robot, which performs multiple plugging and unplugging operations between the charging gun and the vehicle's charging port.

Images