Welding robot pose offset self-correction method, system, device and storage medium

Abstract

The application relates to a welding robot posture offset self-correction method, system, device and storage medium. The method comprises the following steps: time synchronization of multi-source sensor data in the welding robot working process; pre-processing and feature extraction are performed on the synchronized multi-source sensor data to form a multi-dimensional feature vector, and a measurement vector is constructed based on the multi-dimensional feature vector; the confidence of each sensor data is calculated according to the stability index and the residual consistency index, and the measurement noise covariance matrix is determined according to the confidence; based on the measurement vector and the measurement noise covariance matrix, the end posture deviation of the welding robot is fused and estimated to obtain an end posture deviation vector; a trajectory compensation instruction is generated according to the end posture deviation vector, and the welding trajectory is corrected in real time. Through the above scheme, continuous estimation and adaptive compensation of the end posture deviation of the welding robot can be realized under complex welding conditions, and the stability and precision of the welding process are improved.

Description

Technical Field

[0001] This application relates to the field of industrial robot welding technology, specifically to a self-correction method, system, device, and storage medium for welding robot posture deviation. Background Technology

[0002] Currently, welding robots are widely used in automotive and general equipment manufacturing. During the welding process, the robot's end effector center point (TCP) is prone to displacement and attitude angle shift due to factors such as tooling assembly errors, changes in part gaps, thermal deformation, welding torch wear, mechanism compliance and clearance, and external vibrations and collisions. This can lead to weld seam misalignment, fluctuations in forming quality, or even welding failure.

[0003] In related technologies, current welding robot solutions mostly rely on single vision for weld seam tracking or single sensor feedback. On the one hand, visual signals are easily affected by arc light, smoke, and glare, leading to loss or false detection; on the other hand, a single sensor is difficult to stably characterize displacement and attitude angle deviations simultaneously under complex working conditions.

[0004] Therefore, it is necessary to design a new self-correction method for welding robot posture deviation to overcome the above problems. Summary of the Invention

[0005] This application provides a self-correction method, system, device, and storage medium for the attitude deviation of a welding robot, which can solve the technical problems in related technologies where visual signals are easily lost or falsely detected, and a single sensor is difficult to stably characterize displacement and attitude angle deviations simultaneously under complex working conditions.

[0006] In a first aspect, embodiments of this application provide a self-correction method for the posture deviation of a welding robot, the self-correction method for the posture deviation of a welding robot comprising: Time synchronization of multi-source sensor data during the welding robot's operation; Feature extraction is performed on synchronized multi-source sensor data to form multi-dimensional feature vectors. And based on multidimensional feature vectors Construct measurement vector ; Calculate the confidence level of each sensor data based on the signal stability index and residual consistency index of each sensor data. And based on confidence level Determine the measurement noise covariance matrix ; Based on measurement vector With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. ; Based on the end pose deviation vector Generate trajectory compensation instructions, and perform compensation control on the welding trajectory according to the trajectory compensation instructions.

[0007] In conjunction with the first aspect, in one embodiment, the multi-source sensor data includes welding current and voltage signals, welding torch force and torque signals, end effector and gyroscope signals, and joint angle signals.

[0008] In conjunction with the first aspect, in one implementation, the step of extracting features from synchronized multi-source sensor data to form a multi-dimensional feature vector is... And based on multidimensional feature vectors Construct measurement vector ,include: Extracting arc power from welding current and voltage signals Equivalent resistance Steady-state mean and variance, as well as fluctuation and spectral characteristics, are used to construct a current feature vector; Extracting welding normal force from welding torch force and torque signals Direction angle and the rate of change of offset direction And construct the welding torque feature vector; Extracting the welding torch tip angular velocity from end-effector and gyroscope signals. angular acceleration and linear acceleration And calculate the vibration energy index With main frequency Construct inertial measurement feature vectors; Extracting joint angle position from joint angle signals Control command angle Joint velocity and joint acceleration And calculate the joint following error. and Mahal distance Construct joint motion feature vectors; A multidimensional feature vector is constructed based on current feature vector, welding torque feature vector, inertial measurement feature vector, and joint motion feature vector. ; From multidimensional feature vectors The measurement vector is obtained by selecting and concatenating the vectors. .

[0009] In conjunction with the first aspect, in one embodiment, the multi-source sensor data further includes temperature signals and visual and laser sensor signals, and the multidimensional feature vector It also includes temperature feature vectors and visual geometric feature vectors.

[0010] In conjunction with the first aspect, in one implementation, the measurement vector-based... With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. ,include: Establish a state vector, which includes at least the end displacement deviation, the end attitude deviation, and the inertial measurement zero bias; Based on the state vector of the previous time step and the current input vector, the predicted state estimate and predicted covariance of the current time step are obtained using the state prediction model. According to the measurement vector A measurement model is established based on the correspondence between the measurement model and the state vector, and the measurement noise covariance matrix is ​​analyzed based on the confidence level of each sensor data. Perform adaptive updates; Based on predicted state estimates, predicted covariance, and measurement vectors and the updated measurement noise covariance matrix Calculate the Kalman gain and use it to correct the predicted state estimate to obtain the state estimate at the current time. Extract the end-effector displacement deviation and end-effector attitude deviation from the current state estimate, and use them as the end-effector pose deviation vector. .

[0011] In conjunction with the first aspect, in one embodiment, the welding robot attitude offset self-correction method further includes: The compensation range and compensation change rate are constrained, and the welding torch temperature is monitored. When the compensation range, compensation change rate, or welding torch temperature exceeds the limit, the speed is reduced or an alarm is triggered.

[0012] In conjunction with the first aspect, in one embodiment, the method based on the end-effector pose deviation vector... After generating the trajectory compensation instruction, the following is also included: Based on innovative residuals and weld formation quality scoring and arc stability score Confidence level Calculation parameters and measurement noise covariance matrix The mapping parameters are updated.

[0013] Secondly, embodiments of this application also provide a welding robot posture offset self-correction system, the welding robot posture offset self-correction system comprising: The synchronization module is used to synchronize the multi-source sensor data during the welding robot's operation. The feature extraction and measurement construction module is used to extract features from synchronized multi-source sensor data to form multi-dimensional feature vectors. And based on multidimensional feature vectors Construct measurement vector ; The covariance calculation module is used to calculate the confidence level of each sensor data based on the signal stability index and residual consistency index of each sensor data. And based on confidence level Determine the measurement noise covariance matrix ; The fusion estimation module is used for estimation based on measurement vectors. With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. ; The trajectory compensation module is used to calculate the end-effector pose deviation vector. Generate trajectory compensation instructions, and perform compensation control on the welding trajectory according to the trajectory compensation instructions.

[0014] Thirdly, this application also provides a welding robot posture offset self-correction device, which includes a processor, a memory, and a welding robot posture offset self-correction program stored in the memory and executable by the processor. When the welding robot posture offset self-correction program is executed by the processor, it implements the steps of the above-described welding robot posture offset self-correction method.

[0015] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a welding robot attitude offset self-correction program, wherein when the welding robot attitude offset self-correction program is executed by a processor, the steps of the above-described welding robot attitude offset self-correction method are implemented.

[0016] The beneficial effects of the technical solutions provided in this application include: By collecting multi-source sensor data during the welding robot's operation and performing time synchronization and feature extraction on this data, a multi-dimensional feature vector can be constructed. and measurement vector Based on the confidence level of each sensor data The measurement noise covariance matrix can be determined. Based on the measurement noise covariance matrix and measurement vector By performing fusion estimation on the end-effector pose deviation of the welding robot, the end-effector pose deviation vector can be obtained. Furthermore, based on the end-effector pose deviation vector The welding trajectory is corrected and compensated in real time. This embodiment can continuously estimate and compensate for the end attitude error based on multi-source sensor data, get rid of the strong dependence on a single vision, and solve the technical problems in related technologies where visual signals are easily lost or falsely detected and a single sensor is difficult to stably characterize displacement and attitude angle deviations simultaneously under complex working conditions. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an embodiment of the self-correction method for welding robot posture deviation according to this application; Figure 2 This is a schematic diagram of the hardware structure of the self-correction device for welding robot posture deviation involved in the embodiments of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0019] First, some of the technical terms used in this application will be explained to help those skilled in the art understand this application.

[0020] State vector; End displacement deviation; End-point attitude deviation (rotation vector); Zero bias in inertial measurement; : Input vector; Measurement vector; : Feature vector; : State transition function; : Measurement function; Process noise covariance; : Measure noise covariance; Kalman gain; : State transition Jacobi; : Measuring the Jacobian; : Identity matrix; Extremely small amount; Confidence level; Confidence threshold; , , , , Boundary threshold.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0022] In a first aspect, embodiments of this application provide a self-correction method for the posture deviation of a welding robot.

[0023] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the self-correction method for welding robot posture deviation according to this application. Figure 1 As shown, the self-correction method for welding robot attitude deviation includes: S100: Time synchronization of multi-source sensor data during the welding robot's operation.

[0024] S200: Extract features from synchronized multi-source sensor data to form a multi-dimensional feature vector. And based on multidimensional feature vectors Construct measurement vector .

[0025] S300: Calculate the confidence level of each sensor data based on the signal stability index and residual consistency index of each sensor data. And based on confidence level Determine the measurement noise covariance matrix .

[0026] S400: Based on measurement vector With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. .

[0027] S500: Based on the end-effector pose deviation vector Generate trajectory compensation instructions, and perform compensation control on the welding trajectory according to the trajectory compensation instructions.

[0028] It should be understood that prior to step S100, multi-source sensor data during the welding robot's operation will be collected. These multi-source sensors may include current / voltage sensors, a six-dimensional force / torque sensor, an accelerometer and gyroscope (IMU), joint angle sensors, and temperature sensors. The current / voltage sensor reflects changes in the welding circuit load and is used to analyze arc stability and offset trends. The six-dimensional force / torque sensor, mounted at the end of the welding torch or on the flange, monitors welding pressure, normal force fluctuations, and shear force offsets. The accelerometer and gyroscope monitor the vibration and angular acceleration at the end of the welding torch for attitude drift assessment. The joint angle sensor provides the angular position and angular velocity of each joint. The temperature sensor collects the welding torch electrode temperature for thermal drift compensation. Data from each source sensor is collected based on the control cycle, and the timestamps are aligned to obtain a synchronized multi-source data stream.

[0029] In step S100 above, the control cycle is used. (e.g., 1ms to 10ms) Synchronous acquisition , , , , , , , , Etc. Timestamp alignment using hard synchronization or based on network time protocols is employed to form time stamps. The synchronized sample set.

[0030] In step S200, the synchronized multi-source sensor data is preprocessed before feature extraction. Preprocessing includes bandpass / lowpass filtering, peak removal, and drift reduction of the data from each source sensor, followed by Z-score or min-max standardization to obtain a normalized signal, thus reducing the impact of dimensional differences on the fusion estimation. The preprocessing step involves denoising, drift reduction, and standardization of the data from each source.

[0031] This embodiment collects multi-source sensor data during the welding robot's operation and performs time synchronization and feature extraction on the multi-source sensor data to construct a multi-dimensional feature vector. and measurement vector Based on the confidence level of each sensor data The measurement noise covariance matrix can be determined. Based on the measurement noise covariance matrix and measurement vector By performing fusion estimation on the end-effector pose deviation of the welding robot, the end-effector pose deviation vector can be obtained. Then, based on the end pose deviation vector The welding trajectory is corrected and compensated in real time. This embodiment can continuously estimate and compensate for the end attitude error based on multi-source sensor data, get rid of the strong dependence on a single vision, and solve the technical problems in related technologies where visual signals are easily lost or falsely detected and a single sensor is difficult to stably characterize displacement and attitude angle deviations simultaneously under complex working conditions.

[0032] Before step S100, the coordinate system can be defined and calibrated. Establish the robot's base coordinate system. Flange coordinate system Tool coordinate system and sensor coordinate system (Including vision / laser and IMU, etc.). If a vision / laser sensor is configured, extrinsic parameter calibration is performed to obtain the hand-eye calibration matrix. It is used to convert weld geometry measurement results from the sensor system to the robot base coordinate system.

[0033] Furthermore, in one embodiment, the multi-source sensor data includes welding current and voltage signals, welding torch force and torque signals, end effector accelerometer and gyroscope signals, and joint angle signals. Specifically, the welding current and voltage signals are acquired using a current / voltage sensor, the welding torch force and torque signals are acquired using a six-dimensional force / torque sensor, the end effector accelerometer and gyroscope signals are acquired using an accelerometer and gyroscope, and the joint angle signals are acquired using a joint angle sensor.

[0034] Furthermore, in some embodiments, the step of extracting features from the synchronized multi-source sensor data to form a multi-dimensional feature vector is further described. And based on multidimensional feature vectors Construct measurement vector It can include: S201: Extracting arc power from welding current and voltage signals Equivalent resistance We analyze the steady-state mean and variance, as well as the fluctuation and spectral characteristics, and construct a current feature vector.

[0035] S202: Extracting welding normal force from welding torch force and torque signals Direction angle and the rate of change of offset direction And construct the welding torque feature vector.

[0036] S203: Extract the welding torch end angular velocity from the end accelerometer and gyroscope signals. angular acceleration and linear acceleration And calculate the vibration energy index With main frequency Construct inertial measurement feature vectors.

[0037] S204: Extracting joint angle position from joint angle signal Control command angle Joint velocity and joint acceleration And calculate the joint following error. and Mahal distance Construct joint motion feature vectors.

[0038] S205: Construct a multi-dimensional feature vector based on current feature vector, welding torque feature vector, inertial measurement feature vector, and joint motion feature vector. .

[0039] S206: From multidimensional feature vectors The measurement vector is obtained by selecting and concatenating the vectors. .

[0040] In this embodiment, step S201 above extracts electrical parameter features by combining current and voltage signals. With voltage calculate: Arc power: ; Equivalent resistance: ; Steady-state mean and variance: ; Fluctuation characteristics and spectral characteristics: peak-to-peak value, main peak frequency of the spectrum wait.

[0041] The above features are used to characterize arc stability and changes in offset trends. Based on arc power... Equivalent resistance The steady-state mean and variance, along with fluctuation and spectral characteristics, are used to construct the current feature vector. .in, This represents the average welding power. Indicates power variance. Indicates peak power characteristics, This represents the average arc resistance. This represents the variance of the arc resistance.

[0042] In step S202 above, force / torque features are extracted, and the force at the end is collected. With torque Define the transverse force vector. Direction angle:

[0043] Rate of change of offset direction:

[0044] Normal force (welding pressure) (For example, take) (Or the normal component after tool coordinate transformation) is used to characterize the changes in gun tip-workpiece contact / gap and attitude tilt angle, and is used to reduce weight or trigger safety strategies in case of anomalies. Based on welding normal force Direction angle and the rate of change of offset direction Constructing the welding torque eigenvector .in, Indicates the normal force. Indicates the variance of the normal force. Indicates the contact attitude angle. This represents the rate of change of attitude angle.

[0045] In step S203 above, IMU and vibration feature extraction are performed. The angular velocity at the tip of the welding torch is extracted. angular acceleration (Obtained by difference) and linear acceleration Calculate the vibration energy index and dominant frequency: .

[0046] when or When the threshold is exceeded, a chattering / collision condition is determined: on the one hand, deceleration / amplitude limiting is triggered; on the other hand, the relevant measurement noise covariance matrix is ​​increased in the fusion estimation. This enables automatic weight reduction of abnormal signals to prevent filter divergence. It is based on the welding torch tip angular velocity. angular acceleration Vibration energy index With main frequency Constructing inertial measurement feature vectors .in, Indicates angular velocity. Indicates angular acceleration or the rate of change of angular velocity. Indicates the characteristics of vibrational energy. This indicates the characteristic of the dominant frequency.

[0047] In step S204 above, joint and control error features are extracted. Joint angle positions are collected. Control command angle Joint velocity and joint acceleration Calculate the joint following error: .

[0048] The joint angle position is used for positive kinematics calculation of the nominal end-effector pose and serves as a reference for the nominal trajectory; the joint following error is used to reflect the effect of mechanism compliance / backlash on the end-effector pose deviation.

[0049] Then perform anomaly comparison: Generate vectors from key features within the window. Compared with historical normal statistics By comparison, the Mahalanobis distance is obtained: .

[0050] when If an anomaly is detected, a demotion / removal policy or a security policy will be triggered. This is the distance threshold. After calculating the Mahalanobis distance, it is based on the joint velocity. Joint acceleration Joint following error and Mahal distance Constructing joint motion feature vectors .in, This indicates the characteristics of disturbances on the drive side or changes in motor load.

[0051] Furthermore, in one embodiment, the multi-source sensor data further includes temperature signals and visual and laser sensor signals, and the multi-dimensional feature vector It also includes temperature feature vectors and visual geometric feature vectors.

[0052] In this embodiment, the welding robot can be equipped with temperature sensors and vision and laser sensors, such as laser profilometers or industrial cameras, to assist in weld center positioning in low-reflectivity environments. When the confidence level of the visual signal decreases, the system performs deweighting processing on the measurement source to ensure overall estimation stability.

[0053] When configuring vision or laser sensors, the acquired image or contour data is first preprocessed, and the center point or edge features of the weld are extracted to obtain the sensor coordinate system. Coordinates of the center point of the weld below Then, based on the hand-eye calibration obtained from the sensor coordinate system... To the robot's base coordinate system coordinate transformation matrix ,Will Transform to the robot's base coordinate system to obtain ,Right now:

[0054] in, The coordinates of the currently detected weld center point in the robot's base coordinate system are given. Furthermore, the current detected coordinates are compared with the reference coordinates to obtain the geometric deviation of the visual measurement.

[0055] The coordinates of the reference weld center point in the robot's base coordinate system. The geometric deviation... One of the observations that can be used as a visual measurement source is input into the fusion estimation module.

[0056] Visual confidence is obtained by normalizing and weighting the quality and matching stability indices: .

[0057] Where SNR is the signal-to-noise ratio. For the number of effective features, For matching / reprojection errors, blur is the blur index; This is the Sigmoid function.

[0058] In this embodiment, based on geometric deviation Visual confidence Visual geometric feature vectors can be constructed. .

[0059] In this embodiment, the temperature sensor is able to collect temperature data. Based on temperature Constructing temperature feature vectors .

[0060] To improve the efficiency of utilizing multi-source heterogeneous information in attitude offset estimation for welding robots, this embodiment simultaneously acquires, preprocesses, and extracts features from current, torque, inertial, joint, temperature, and visual signals to construct a multi-dimensional feature vector. This vector is then further selected, concatenated, or mapped to form an extended Kalman filter (EKF) measurement vector. The desired features are then selected and concatenated sequentially to obtain the EKF measurement vector. .

[0061] The above multidimensional feature vectors It can contain temperature feature vectors and visual geometric feature vectors That is, multidimensional feature vectors .

[0062] In step S200 above, features such as electrical parameters, force / torque, IMU, joint and temperature are extracted using a sliding window to form a multidimensional feature vector. And construct the measurement vector accordingly. .

[0063] Furthermore, in one embodiment, step S300 above mainly focuses on the confidence level of non-visual sources. Calculate the confidence level of the non-visual source. For sources such as electrical parameters, force / torque, IMU, and joints, a combined calculation of "stability + residual consistency" is used: .in For key features (such as) , , wait), This is the normalized square of innovation (calculated from EKF innovation).

[0064] In step S300, the mapping from confidence level to weight / noise covariance adopts an engineering-equivalent and easily fusion-friendly method, whereby the weights are reflected by adjusting the measurement noise covariance matrix: Measurement noise covariance matrix Calculate the confidence level for each measurement source (i.e., sensor). Map the confidence level to the measurement noise covariance matrix (or equivalent weight) to achieve dynamic weight reduction / removal.

[0065] when When the confidence threshold is reached, remove the measurement source or set... Approximate elimination is achieved by taking the maximum value. The lower the confidence level, the better. The larger the value, the more automatically the filter will reduce the weight of that source.

[0066] Furthermore, in one embodiment, the measurement vector-based... With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. ,include: S401: Establish the state vector. In this embodiment, the state vector includes at least the end-effector displacement deviation, the end-effector attitude deviation, and the inertial measurement zero bias.

[0067] S402: State Prediction. Based on the state vector from the previous time step and the current input vector, the predicted state estimate and predicted covariance at the current time step are obtained using the state prediction model.

[0068] S403: Establish measurement mapping relationships and update the measurement noise covariance matrix. Establish a measurement model based on the correspondence between measurement vectors and state vectors, and adaptively update the measurement noise covariance matrix according to the confidence level of each sensor data.

[0069] S404: State Correction. Calculate the Kalman gain based on the predicted state estimate, prediction covariance, measurement vector, and updated measurement noise covariance matrix. Use the Kalman gain to correct the predicted state estimate to obtain the state estimate at the current time.

[0070] S405: Output end-effector pose deviation vector. Extract the end-effector displacement deviation and end-effector attitude deviation from the current state estimate and output them as the end-effector pose deviation vector.

[0071] In this embodiment, EKF (Extended Kalman Filter) can be used to predict and update the end-effector pose deviation state, resulting in the end-effector pose deviation vector. .

[0072] The specific steps of fusion estimation include: Step 1: State Definition Define the state vector: .

[0073] in, For end displacement deviation, This represents the end-effector attitude deviation (rotation vector / small angle). and The IMU zero bias is used for drift compensation.

[0074] Step 2: Prediction Model .

[0075] in, is the input vector, representing the system input vector at time k, which consists of joint commands, joint measurements, and related control inputs, and is used for nominal kinematic propulsion and deviation propagation; Let be the process noise vector, representing the process noise vector at time k, used to characterize the modeling error and external disturbances during the system state propagation process; Let be the process noise covariance.

[0076] Step 3: Measurement Model .

[0077] in, Depend on Select / assemble to obtain, From each source It is composed and adaptively updated according to confidence level; The measurement noise vector is denoted by k, which represents the measurement noise vector at time k and is used to characterize random noise and observation error in the multi-sensor measurement process.

[0078] Step 4: EKF update formula: Predicting covariance: ,in, .

[0079] Kalman gain: ,in, .

[0080] Status Update: .

[0081] Covariance update: .

[0082] In this embodiment, the gain matrix Determine the "magnitude of the measurement's correction to the state". When the confidence level of a measurement source is high (corresponding to...) Hour, Increasing the confidence level makes the filter more confident in the measurement source; when the confidence level is low... The size is reduced, thereby enabling adaptive fusion and avoiding estimation divergence caused by anomalous noise.

[0083] Step 5: End-effector pose deviation vector output and trajectory compensation. Map the state estimation vector output by the EKF to the end-effector pose deviation vector:

[0084] in, Indicates the displacement deviation at the end point. Represents the end-effector attitude deviation (in rotation vector form). This represents the mapping from rotation matrix to rotation vector.

[0085] The nominal pose is: , .

[0086] end pose deviation vector After being converted into compensation commands and superimposed onto the nominal trajectory, the compensated control commands are obtained:

[0087] If the controller uses joint space compensation, the joint compensation is obtained through the pseudo-inverse mapping of the Jacobian matrix J: in, It is a pseudo-inverse matrix.

[0088] Furthermore, in one embodiment, the self-correction method for welding robot posture deviation further includes: constraining the compensation amplitude and the compensation change rate, and monitoring the welding torch temperature; and reducing speed or issuing an alarm when the compensation amplitude, the compensation change rate, or the welding torch temperature exceeds the limit.

[0089] This embodiment sets three types of boundaries for the compensation command to impose safety constraints, including: 1. Amplitude saturation: , ; 2. Rate of change limit: This embodiment constrains the linear velocity and angular velocity for end-effector pose compensation: .

[0090] in, and These represent the end displacement deviation and attitude deviation at time k, respectively. The sampling period is For the maximum permissible end linear velocity, This is the maximum permissible end-effector angular velocity. When the calculated end-effector compensation velocity or angular velocity exceeds the above threshold, the controller will automatically reduce the compensation gain or trigger a limiting action to ensure a smooth welding process.

[0091] 3. Thermal protection: When the temperature... When the thermal protection threshold is reached, the compensation gain is reduced, process noise is increased, and a speed reduction / alarm is triggered, while the thermal drift compensation item is enabled. .

[0092] Furthermore, in one embodiment, the drift compensation mechanism includes: IMU zero-bias drift: Incorporate online state estimation and bias correction: , .

[0093] Thermal drift: establishing a mapping based on temperature It can be added or deducted as a priori compensation item.

[0094] Slow zero drift: During the steady-state operating window Low-pass filtering and zero-point calibration suppress long-term cumulative deviations.

[0095] Furthermore, in one embodiment, in the step based on the end-effector pose deviation vector After generating the trajectory compensation instruction, it also includes: scoring based on innovative residuals and weld formation quality. and arc stability score Confidence level Calculation parameters and measurement noise covariance matrix The mapping parameters are updated.

[0096] In this embodiment, the self-calibration method adopts a closed loop of "feedforward compensation + feedback self-learning".

[0097] Feedforward: The compensation for the next cycle is directly generated and superimposed on the trajectory command to quickly offset the offset.

[0098] Feedback: Scoring based on innovation residuals and weld formation quality. Arc stability score Update the fusion parameters.

[0099] Among them, the predicted offset: .

[0100] Actual offset The method of obtaining it is: 1. When visual / laser confidence At that time, with (and optional attitude deviation) as ; 2. When vision is unavailable, provide periodic measurement results based on post-weld sampling / calibration workpiece measurements. Online, innovative residuals and process indicators serve as weak supervision constraints.

[0101] The self-learning update objects in this embodiment include at least: , and confidence mapping parameters (such as) , ).renew , , This will directly affect the innovation covariance and Kalman gain. This, in turn, affects state estimation. The convergence speed and the degree of trust in each source are used to achieve adaptive operation under different working conditions.

[0102] This application can break free from the strong dependence on a single vision system and estimate the end effector pose deviation by fusing information from multiple sources such as electrical parameters, force / torque, IMU, joint signals, and temperature. Through a confidence / weight adaptive mechanism, it automatically reduces or eliminates signals from each sensor source when the signal quality changes, avoiding filter divergence. Furthermore, it converts the estimated displacement error and attitude angle error into compensation commands and superimposes them onto the trajectory in real time to achieve feedforward compensation. It also introduces drift compensation and closed-loop self-learning to adaptively update the noise covariance and weight mapping parameters, improving long-term stability and cross-workpiece adaptability.

[0103] Secondly, embodiments of this application also provide a self-correction system for the posture deviation of a welding robot.

[0104] In one embodiment, the welding robot attitude deviation self-correction system includes: a synchronization module for time synchronization of multi-source sensor data during the welding robot's operation; and a feature extraction and measurement construction module for extracting features from the synchronized multi-source sensor data to form a multi-dimensional feature vector. And based on multidimensional feature vectors Construct measurement vector The covariance calculation module is used to calculate the confidence level of each sensor data based on the signal stability index and residual consistency index of each sensor data. And based on confidence level Determine the measurement noise covariance matrix ; Fusion estimation module, which is used for estimation based on measurement vectors With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. The trajectory compensation module is used to calculate the end-effector pose deviation vector. Generate trajectory compensation instructions, and perform compensation control on the welding trajectory according to the trajectory compensation instructions.

[0105] In this embodiment, the welding robot posture offset self-correction system also includes the robot body. The aforementioned synchronization module, feature extraction and measurement construction module, covariance calculation module, fusion estimation module, and trajectory compensation module are all installed on the robot body. In the above embodiment, the feature extraction and measurement construction module constructs feature vectors and measurement vectors from each source signal. The fusion estimation module can estimate the end-effector pose deviation based on the extended Kalman filter (EKF) or an equivalent nonlinear fusion filtering framework. The trajectory compensation module maps the deviation to end-effector pose compensation or joint compensation and superimposes it onto the nominal trajectory.

[0106] The functions of each module in the above-mentioned welding robot attitude offset self-correction system correspond to the steps in the above-mentioned welding robot attitude offset self-correction method embodiment, and their functions and implementation processes will not be described in detail here.

[0107] In some optional embodiments, the above-mentioned welding robot posture deviation self-correction system further includes a multi-sensor unit, which may include a current / voltage sensor, a six-dimensional force / torque sensor, an accelerometer and gyroscope, and a joint encoder. The current / voltage sensor can collect welding current and voltage signals, the six-dimensional force / torque sensor can collect welding torch force and torque signals, the accelerometer and gyroscope can collect end-effector accelerometer and gyroscope signals, and the joint angle sensor can collect joint angle signals.

[0108] The multi-sensor unit may also include a temperature sensor and vision and laser sensors. The temperature sensor can acquire temperature signals, and the vision and laser sensors can acquire vision and laser signals. The vision and laser sensors are used for weld geometry positioning.

[0109] Furthermore, in one embodiment, the welding robot attitude deviation self-correction system also includes a safety strategy module, which is used to constrain the compensation amplitude and compensation rate of change, and monitor the welding torch temperature. When the compensation amplitude, compensation rate of change, or welding torch temperature exceeds the limit, the system will reduce speed or issue an alarm. This embodiment constrains the compensation amplitude, rate of change (linear velocity / angular velocity), and thermal protection.

[0110] Furthermore, in one embodiment, the welding robot posture offset self-correction system also includes a closed-loop feedback and self-learning module, which is used for scoring based on innovative residuals and weld formation quality. and arc stability score Confidence level Calculation parameters and measurement noise covariance matrix The mapping parameters are updated. In this embodiment, the weights and noise covariance are updated based on the residuals and welding quality / arc stability indices to achieve long-term adaptive behavior.

[0111] This application implements multi-source redundancy, enabling offset estimation through electrical parameters, forces, IMU, and joints even when vision is unavailable; it exhibits adaptive robustness, with confidence-driven noise covariance updates achieving dynamic weighting to prevent filter divergence caused by anomalous signals; it is highly implementable, providing explicit modeling, measurement construction, compensation mapping, threshold boundaries, and symbol definitions; and it demonstrates long-term stability by introducing drift compensation and closed-loop self-learning to reduce the impact of thermal drift / zero bias accumulation on bias estimation.

[0112] Thirdly, embodiments of this application provide a self-correction device for the posture deviation of a welding robot. The self-correction device for the posture deviation of a welding robot can be a device with data processing capabilities, such as a personal computer (PC), a laptop computer, or a server.

[0113] Reference Figure 2 , Figure 2 This is a schematic diagram of the hardware structure of the welding robot posture offset self-correction device involved in the embodiments of this application. In the embodiments of this application, the welding robot posture offset self-correction device may include a processor, a memory, a communication interface, and a communication bus.

[0114] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.

[0115] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal components of the welding robot's attitude offset self-correction device, as well as interfaces used for interconnecting the welding robot's attitude offset self-correction device with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.

[0116] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0117] The processor can be a general-purpose processor, which can call the welding robot posture offset self-correction program stored in the memory and execute the welding robot posture offset self-correction method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the welding robot posture offset self-correction program is called can refer to the various embodiments of the welding robot posture offset self-correction method of this application, and will not be repeated here.

[0118] Those skilled in the art will understand that Figure 2 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0119] Fourthly, embodiments of this application also provide a readable storage medium.

[0120] The present application has a readable storage medium storing a welding robot attitude offset self-correction program, wherein when the welding robot attitude offset self-correction program is executed by a processor, the steps of the welding robot attitude offset self-correction method as described above are implemented.

[0121] The method implemented when the welding robot posture offset self-correction program is executed can be referred to in various embodiments of the welding robot posture offset self-correction method of this application, and will not be repeated here.

[0122] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0123] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.

[0124] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.

[0125] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0126] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0127] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.

[0128] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A self-correction method for attitude deviation of a welding robot, characterized in that, The self-correction method for welding robot posture deviation includes: Time synchronization of multi-source sensor data during the welding robot's operation; Feature extraction is performed on synchronized multi-source sensor data to form multi-dimensional feature vectors. And based on multidimensional feature vectors Construct measurement vector ; Calculate the confidence level of each sensor data based on the signal stability index and residual consistency index of each sensor data. And based on confidence level Determine the measurement noise covariance matrix ; Based on measurement vector With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. ; Based on the end pose deviation vector Generate trajectory compensation instructions, and perform compensation control on the welding trajectory according to the trajectory compensation instructions.

2. The self-correction method for welding robot attitude deviation as described in claim 1, characterized in that, The multi-source sensor data includes welding current and voltage signals, welding torch force and torque signals, end effector and gyroscope signals, and joint angle signals.

3. The self-correction method for welding robot attitude deviation as described in claim 1, characterized in that, The synchronized multi-source sensor data is used to extract features to form a multi-dimensional feature vector. And based on multidimensional feature vectors Construct measurement vector ,include: Extracting arc power from welding current and voltage signals Equivalent resistance Steady-state mean and variance, as well as fluctuation and spectral characteristics, are used to construct a current feature vector; Extracting welding normal force from welding torch force and torque signals Direction angle and the rate of change of offset direction And construct the welding torque feature vector; Extracting the welding torch tip angular velocity from end-effector and gyroscope signals. angular acceleration and linear acceleration And calculate the vibration energy index With main frequency Construct inertial measurement feature vectors; Extracting joint angle position from joint angle signals Control command angle Joint velocity and joint acceleration And calculate the joint following error. and Mahal distance Construct joint motion feature vectors; A multidimensional feature vector is constructed based on current feature vector, welding torque feature vector, inertial measurement feature vector, and joint motion feature vector. ; From multidimensional feature vectors The measurement vector is obtained by selecting and concatenating the vectors. .

4. The self-correction method for welding robot attitude deviation as described in claim 3, characterized in that, The multi-source sensor data also includes temperature signals and visual and laser sensor signals, and the multi-dimensional feature vector... It also includes temperature feature vectors and visual geometric feature vectors.

5. The self-correction method for welding robot attitude deviation as described in claim 1, characterized in that, The measurement vector-based With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. ,include: Establish a state vector, which includes at least the end displacement deviation, the end attitude deviation, and the inertial measurement zero bias; Based on the state vector of the previous moment and the current input vector, the predicted state estimate and predicted covariance of the current moment are obtained using the state prediction model. According to the measurement vector A measurement model is established based on the correspondence between the measurement model and the state vector, and the measurement noise covariance matrix is ​​analyzed based on the confidence level of each sensor data. Perform adaptive updates; Based on predicted state estimates, predicted covariance, and measurement vectors and the updated measurement noise covariance matrix Calculate the Kalman gain and use it to correct the predicted state estimate to obtain the state estimate at the current time. Extract the end-effector displacement deviation and end-effector attitude deviation from the current state estimate, and use them as the end-effector pose deviation vector. .

6. The self-correction method for welding robot attitude deviation as described in claim 1, characterized in that, The self-correction method for welding robot posture deviation also includes: The compensation range and compensation change rate are constrained, and the welding torch temperature is monitored. When the compensation range, compensation change rate, or welding torch temperature exceeds the limit, the speed is reduced or an alarm is triggered.

7. The self-correction method for welding robot attitude deviation as described in claim 1, characterized in that, According to the end pose deviation vector After generating the trajectory compensation instruction, the following is also included: Based on innovative residuals and weld formation quality scoring and arc stability score Confidence level Calculation parameters and measurement noise covariance matrix The mapping parameters are updated.

8. A self-correction system for attitude deviation of a welding robot, characterized in that, The welding robot attitude offset self-correction system includes: The synchronization module is used to synchronize the multi-source sensor data during the welding robot's operation. The feature extraction and measurement construction module is used to extract features from synchronized multi-source sensor data to form multi-dimensional feature vectors. And based on multidimensional feature vectors Construct measurement vector ; The covariance calculation module is used to calculate the confidence level of each sensor data based on the signal stability index and residual consistency index of each sensor data. And based on confidence level Determine the measurement noise covariance matrix ; The fusion estimation module is used for estimation based on measurement vectors. With measurement noise covariance matrix The end-effector pose deviation of the welding robot is fused and estimated to obtain the end-effector pose deviation vector. ; The trajectory compensation module is used to calculate the end-effector pose deviation vector. Generate trajectory compensation instructions, and perform compensation control on the welding trajectory according to the trajectory compensation instructions.

9. A self-correction device for the attitude deviation of a welding robot, characterized in that, The welding robot attitude offset self-correction device includes a processor, a memory, and a welding robot attitude offset self-correction program stored in the memory and executable by the processor, wherein when the welding robot attitude offset self-correction program is executed by the processor, it implements the steps of the welding robot attitude offset self-correction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a welding robot attitude offset self-correction program, wherein when the welding robot attitude offset self-correction program is executed by a processor, it implements the steps of the welding robot attitude offset self-correction method as described in any one of claims 1 to 7.

Images