Arc welding task-oriented embodied robot multi-sensor position self-calibration method
By constructing a propagation distortion field and distortion compensation model, the problem of position calibration distortion caused by propagation distortion during arc welding was solved, which improved the accuracy and stability of the self-calibration of the multi-sensor position of the embodied robot and enhanced its adaptive capability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI GUANZHI IND AUTOMATION
- Filing Date
- 2026-04-28
- Publication Date
- 2026-06-05
Smart Images
Figure CN122143045A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of robot arc welding position calibration technology, and in particular to a multi-sensor position self-calibration method for embodied robots for arc welding operations. Background Technology
[0002] Embossed robots have been gradually applied to arc welding operations on boxes, plates, pipes, and complex curved surface components. During arc welding, the robot usually needs to simultaneously complete the adjustment of the welding torch position, the recognition of workpiece features, the tracking of the weld position, and the trajectory correction. Therefore, the accuracy of the spatial positional relationship between the welding torch, the workpiece, and the weld affects the weld formation quality and stability.
[0003] Existing technologies typically employ visual sensors, laser sensors, distance sensors, or acoustic sensors to acquire spatial position data of the welding torch, workpiece, and weld seam, combining offline calibration, online compensation, or multi-sensor fusion methods to achieve position correction. While these methods can achieve some success in static environments, they generally assume that the propagation space between the sensor and the target is stable, and that the acquired observations accurately reflect the target's position.
[0004] However, in actual arc welding, a hot plume continuously forms in front of the welding torch and above the workpiece, accompanied by shielding gas flow disturbance, metal vapor diffusion, and fume accumulation. These factors alter the propagation time, intensity, and direction of the sensor signal in the near-arc region, causing dynamic shifts in the observation position. Existing multi-sensor position calibration schemes primarily correct for workpiece installation deviations, welding torch posture changes, and sensor installation errors, lacking effective modeling and compensation for near-arc region propagation distortion, leading to easily distorted and drifting position calibration results.
[0005] Therefore, this invention proposes a multi-sensor position self-calibration method for an embodied robot for arc welding operations. The information disclosed in the background section is only for enhancing understanding of the background of this disclosure and may therefore contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention
[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a multi-sensor position self-calibration method for embodied robots used in arc welding operations, thereby solving the technical problems mentioned in the background section.
[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for multi-sensor position self-calibration of an embodied robot for arc welding operations includes the following steps: S1. Read the Beidou positioning output, robot end nominal pose, workpiece nominal reference position, sensor installation pose, historical calibration parameters and arc-free reference path set corresponding to the welding task, collect and complete the time alignment and spatial alignment of the original multipath measurement records, and obtain the near-arc area multipath observation dataset. S2. Based on the near-arc region multipath observation dataset and the arc-free reference path set, the arrival time deviation, propagation intensity attenuation and direction deflection are extracted to generate an effective path feature set, and the initial distortion distribution results are formed according to spatial units. S3. Using the initial distortion distribution results as the initial solution, the particle swarm optimization algorithm is used to solve the target propagation distortion field, and an observation distortion correspondence set and distortion compensation model are established. S4. Based on the distortion compensation model, reverse correction and fusion are performed on the welding gun position, workpiece feature position and weld position to obtain a unified correction position set and a high conflict record set, and a calibration constraint set is generated. S5. Update the center point parameters of the welding torch tool, sensor extrinsic parameters and robot end pose mapping relationship according to the calibration constraint set, and output the position self-calibration result after back-substitution verification.
[0008] S1 specifically includes: reading the BeiDou positioning output corresponding to the welding task, the nominal pose of the robot end effector, the nominal reference position of the workpiece, the installation pose of each sensor, historical calibration parameters, and the arc-free reference path set to form initial spatiotemporal reference data; controlling multiple sensors to synchronously transmit, receive, or sample under a unified time reference based on the initial spatiotemporal reference data to generate an original multipath measurement record set containing propagation path identifiers and corrected acquisition times; performing time alignment, spatial alignment, and observation unit merging on the original multipath measurement record set in conjunction with the arc-free reference path set, deleting observation units without reference paths or with insufficient number of effective propagation paths, and outputting a near-arc region multipath observation dataset.
[0009] S2 specifically includes: reading the near-arc region multipath observation dataset and the arc-free reference path set; performing path identification pairing verification on each propagation path; deleting propagation path records that cannot be matched, whose direction vectors cannot be constructed, or whose fields are missing; and calculating the arrival time deviation, propagation intensity attenuation, and direction deflection to generate an effective path feature set; based on the effective path feature set and the aligned spatial position of each propagation path, mapping each path feature to the corresponding spatial unit; statistically analyzing the time deviation, propagation intensity attenuation, and direction deflection of each spatial unit to form an initial distortion distribution result; using the initial distortion distribution result as the initial solution, iteratively updating the time distortion parameters, intensity distortion parameters, and direction distortion parameters of each spatial unit using the particle swarm optimization algorithm to obtain the target propagation distortion field.
[0010] S3 specifically includes: reading the target propagation distortion field, initial spatiotemporal reference data, absolute reference coordinates of the workstation, nominal reference position of the workpiece, and actual pose of the robot end effector; performing coordinate transformation, task consistency verification, and deletion of invalid spatial units on the spatial unit positions to form distortion field registration base data; determining the spatial units traversed by the propagation path, the action order, and the action weight based on the distortion field registration base data; establishing a set of observation distortion correspondences between the observation source, the observation object, and the distortion parameters of the spatial units; generating coordinate correction vectors for each spatial unit according to the observation distortion correspondence set; and establishing a distortion compensation model for transforming the observation coordinates to the real spatial coordinates according to the action order and action weight.
[0011] S4 specifically includes: reading the distortion compensation model, performing compensation correction on the welding torch position observation values, workpiece feature position observation values, and weld seam position observation values output by each sensor, deleting records that exceed the workpiece boundary, welding torch operation boundary, or compensation residual limit after correction, and generating a set of corrected position records; grouping and fusing the set of corrected position records according to the observation object and sampling time to obtain a unified set of corrected positions, and marking the set of high-conflict records according to the distance difference between the multi-source corrected positions; based on the unified set of corrected positions, the set of high-conflict records, the nominal pose of the robot end effector, and historical calibration parameters, solving for the offset of the welding torch tool center point, the offset of the sensor extrinsic parameters, and the offset of the robot end effector pose mapping, forming a calibration constraint set.
[0012] S5 specifically includes: reading the calibration constraint set, distortion compensation model, unified correction position set, and historical calibration parameters; splitting parameter offsets according to parameter type; and updating the welding torch tool center point parameters, sensor extrinsic parameters, and robot end-effector pose mapping relationship according to constraint reliability to generate an updated calibration parameter set; substituting the updated calibration parameter set back into the distortion compensation model to form a predicted position set; comparing it with the unified correction position set to obtain the remaining deviation and parameter verification results; and selecting the calibration parameter set that passes verification; generating position self-calibration results based on the verified calibration parameter set and parameter verification results, and writing them into the robot control system and historical calibration parameter library.
[0013] The beneficial effects of this invention are as follows: This invention extracts arrival time deviation, propagation intensity attenuation, and direction deflection by reading BeiDou positioning output, arc-free reference path set, and near-arc multipath observation dataset, and constructs a target propagation distortion field accordingly. This enables explicit characterization of propagation disturbances caused by hot plume, shielding gas flow, and smoke during arc welding, and solves the position calibration distortion problem caused by the prior art's default assumption that the propagation space is a stable space.
[0014] This invention establishes a set of observation distortion correspondences and a distortion compensation model based on the target propagation distortion field, and performs reverse correction on the welding torch position, workpiece feature position, and weld position. This converts observation coordinates affected by near-arc region propagation distortion into true spatial coordinates, improving the accuracy of multi-sensor position self-calibration results. By performing unified fusion of multi-source calibration positions and adjusting constraint reliability using a high-conflict record set, the invention suppresses the biasing effect of abnormal observation sources and conflicting observations on the calibration results, improving the consistency and stability of multi-source observation fusion results.
[0015] This invention solves for the offset of the welding torch tool center point, the offset of sensor extrinsic parameters, and the offset of robot end-effector pose mapping based on a unified calibration position set. It then performs updates and back-substitution verification on these three types of parameters, enabling closed-loop correction of position calibration parameters and improving the adaptive capability of the robot in continuous arc welding operations. The verified calibration parameters are written into the robot control system and the historical calibration parameter library, allowing the current position self-calibration results to be directly used for subsequent arc welding trajectory correction, weld repositioning, and recall in the next sampling cycle. This improves the timeliness of position calibration and the reliability of continuous operation throughout the entire arc welding process. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of a multi-sensor position self-calibration method for a holographic robot for arc welding operations according to the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Example: Figure 1 As shown, this embodiment provides a multi-sensor position self-calibration method for a holographic robot for arc welding operations, including the following steps: S1. Read the Beidou positioning output, robot end nominal pose, workpiece nominal reference position, sensor installation pose, historical calibration parameters and arc-free reference path set corresponding to the welding task, collect and complete the time alignment and spatial alignment of the original multipath measurement records, and obtain the near-arc area multipath observation dataset. S2. Based on the near-arc region multipath observation dataset and the arc-free reference path set, the arrival time deviation, propagation intensity attenuation and direction deflection are extracted to generate an effective path feature set, and the initial distortion distribution results are formed according to spatial units. S3. Using the initial distortion distribution results as the initial solution, the particle swarm optimization algorithm is used to solve the target propagation distortion field, and an observation distortion correspondence set and distortion compensation model are established. S4. Based on the distortion compensation model, reverse correction and fusion are performed on the welding gun position, workpiece feature position and weld position to obtain a unified correction position set and a high conflict record set, and a calibration constraint set is generated. S5. Update the center point parameters of the welding torch tool, sensor extrinsic parameters and robot end pose mapping relationship according to the calibration constraint set, and output the position self-calibration result after back-substitution verification.
[0019] S1 specifically includes: S110. In this embodiment, the welding task identifier, robot end nominal pose, workpiece nominal reference position, sensor installation pose, historical calibration parameters, and Beidou positioning output corresponding to the current welding task are read first, and the integrity of the read content is verified. Among them, the Beidou positioning output is defined as a data set used to characterize the absolute position of the welding station and the unified timing result. Its purpose is to provide the same time reference and the same station reference for multi-sensor collaborative acquisition, so as to avoid the incomparability of subsequent measurement records caused by local clock drift or station switching of each sensor.
[0020] Data records with missing fields, mismatched task identifiers, or missing workstation reference coordinates are deleted, retaining only data items consistent with the current welding task. After verification, the initial spatiotemporal reference data and the arc-free reference path set are output for use by S120 and S130.
[0021] The initial spatiotemporal reference data includes at least: welding task identifier, absolute reference coordinates of the workstation, unified time synchronization, nominal pose of the robot end effector, nominal reference position of the workpiece, installation pose of each sensor, and historical calibration parameters. The historical calibration parameters include at least the historical values of the welding torch tool center point parameters, the historical values of the sensor extrinsic parameters, and the historical values of the robot end effector pose mapping. The purpose of setting these parameters is to provide a preliminary constraint basis for subsequent propagation distortion field registration and calibration parameter updates. The historical calibration parameters are stored in the historical calibration parameter library. The historical calibration parameters read subsequently all refer to the valid historical calibration parameters read from the historical calibration parameter library and verified and retained.
[0022] The arc-free reference path set is defined as a set of reference propagation path records that are pre-established and stored in association with the current welding task under the arc-stopped state. It includes at least the path identifier, arc-free arrival time, arc-free propagation intensity, and arc-free propagation direction. Its purpose is to provide a unified comparison benchmark for subsequent calculation of the deviation of each propagation path.
[0023] For example, for welding tasks with the same workstation number 03, the arc-free reference paths identified as P1, P2, and P3 can be pre-recorded, where the arc-free arrival time of P1 can be recorded as 12.4 microseconds. Subsequently, only the current propagation paths corresponding one-to-one with P1, P2, and P3 will be included in the processing to ensure comparability from the source.
[0024] S120: Read the unified time synchronization, sensor installation pose and robot end nominal pose from the initial spatiotemporal reference data, control the multiple sensors deployed around the welding torch and the workstation to perform synchronous transmission, synchronous reception or synchronous sampling under the unified time reference, obtain welding torch position observation value, workpiece feature position observation value and weld position observation value, and generate original multipath measurement record.
[0025] After generating the original multipath measurement records, the time of each record is first corrected, and then the corrected records are verified to meet the acquisition requirements under the same welding task. Records that do not meet the requirements are deleted, and finally the original multipath measurement record set is output for S130 to call.
[0026] In this embodiment, each original multipath measurement record includes at least the propagation path identifier, launch time, reception time, propagation intensity, incident direction, reception direction, the actual pose of the robot's end effector at the corresponding sampling time, and the acquisition time corrected according to a unified time reference. The time correction is performed according to the following formula: in, For the first The corrected acquisition time of each measurement record; For the first The original acquisition time of each measurement record; The first time reference obtained based on the unified time reference of Beidou positioning The measurement record clock correction amount. The purpose of the above processing is to unify the local acquisition time of different sensors to the same time reference, so that subsequent judgments for the same sampling period have a consistent basis. Measurement records with missing transmission time, missing reception time, missing propagation path identification, missing actual pose of the robot end effector, or whose corrected acquisition time exceeds the time range of the current welding task are deleted.
[0027] For example, if the original acquisition time of a record is 25.000 milliseconds and the corresponding clock correction is 0.012 milliseconds, its corrected acquisition time will be 25.012 milliseconds. The record will only be retained for subsequent processing if the corrected acquisition time falls within the sampling window of the current welding task. After the above processing, each record in the original multipath measurement record set will be comparable under a unified time reference.
[0028] S130: Read the sensor installation poses from the arc-free reference path set and the initial spatiotemporal reference data, as well as the original multipath measurement record set and the actual poses of the robot end effector contained therein. Perform time alignment, spatial alignment, and observation unit merging on the original multipath measurement record set in sequence. After merging, delete the observation units that do not meet the requirements of subsequent propagation difference analysis, and output the near-arc region multipath observation dataset for S210 to call.
[0029] Time alignment refers to classifying adjacent measurement records into the same sampling period based on the corrected acquisition time. This embodiment adopts the following determination rule: in, For the first The corrected acquisition time of each path record; For the first The corrected acquisition time of each path record; The time window threshold for the same sampling period; This represents the number of effective propagation paths within the observation unit. Time window threshold. The purpose of this setting is to limit the time discrete range of each path record within the same sampling period, for example, it can be 0.05 milliseconds; when the difference between the corrected acquisition times of two path records does not exceed this threshold, they are considered to be in the same sampling period.
[0030] Spatial alignment refers to uniformly transforming the measurement positions of each sensor in its local coordinate system to the same reference coordinate system. The transformation relationship is performed according to the following formula: in, For the first The position vector of each measurement record in a unified reference coordinate system; For the first The position vector of each measurement record in the corresponding sensor's local coordinate system; For the first The coordinate rotation matrix corresponding to each measurement record; For the first The coordinate translation vector corresponding to each measurement record.
[0031] The purpose of the aforementioned spatial alignment is to eliminate differences in the installation orientation and position of different sensors, so that subsequent propagation paths can establish a corresponding relationship under the same spatial reference.
[0032] After completing time and spatial alignment, path records belonging to the same sampling period, located in the same near-arc region, and capable of establishing a one-to-one correspondence with the same path identifier records in the non-arc reference path set are grouped into the same observation unit. Observation units that cannot find a reference path, have mismatched task identifiers, or have fewer than 2 effective propagation paths are deleted. The reason is that when there are fewer than 2 paths, it is impossible to form the minimum comparison basis for subsequent propagation time deviation, propagation intensity attenuation, and direction deflection.
[0033] The final output of the near-arc region multipath observation dataset includes at least the observation unit identifier, path identifier set, aligned acquisition time, aligned spatial position, corresponding reference path identifier, and corresponding robot end-effector actual pose. This output will be directly used to extract the effective path feature set in subsequent steps.
[0034] S2 specifically includes: S210. Read each observation unit in the near-arc region multipath observation dataset and simultaneously read the reference path records corresponding to each propagation path identifier in the arc-free reference path set. Perform pairing verification on each propagation path in each observation unit. The verification content includes at least whether the path identifier is consistent, whether the current sampling time exists, whether the propagation intensity field exists, and whether the current propagation direction vector can be determined by the incident direction and the receiving direction.
[0035] For propagation path records that cannot establish a one-to-one correspondence with the arc-free reference path set, propagation path records whose propagation direction vectors cannot be constructed, and propagation path records with missing fields, deletion is performed. After retaining comparable propagation path records, the arrival time deviation, propagation intensity attenuation, and direction deflection are calculated, and the effective path feature set is output for S220 and S230 to call.
[0036] The arrival time deviation characterizes the change in propagation time between the current near-arc region propagation state and the arc-stopping reference state; the propagation intensity attenuation characterizes the degree to which the near-arc region's thermal plume, protective airflow, and dust weaken the propagation energy; and the directional deflection characterizes the degree of directional shift of the propagation path under disturbance. These three characteristics are calculated using the following formula: in, For the first The arrival time deviation of each propagation path; For the first The arrival time of each propagation path in the current sampling period; For the first The reference arrival time of each propagation path in the arc-free reference path set; For the first The attenuation of propagation intensity along each propagation path; For the first The propagation intensity of each propagation path in the current sampling period; For the first The reference propagation intensity of each propagation path in the arc-free reference path set; For the first The directional deflection of the propagation path; For the first The propagation direction vector of each propagation path in the current sampling period; For the first The reference propagation direction vector of each propagation path in the set of arc-free reference paths. Wherein, " represents the vector dot product operation, " represents the vector magnitude operation, "" indicates the inverse cosine function operation. To avoid outliers from entering subsequent spatial statistics, ... Exceeding the preset physical limit Less than 0 or greater than 1, and Path feature records exceeding the preset angle limit will be deleted.
[0037] For example, if the reference propagation intensity of a certain path is 10 and the current propagation intensity is 7, then the propagation intensity attenuation is 0.3, and this record can be directly retained into the effective path feature set.
[0038] S220: Read the effective path feature set and read the aligned spatial positions of each propagation path in the near-arc zone multi-path observation dataset. Using the current spatial position of the welding torch as the center and the nominal reference position of the workpiece as the boundary of the near-arc zone, establish a set of spatial cells to characterize the local disturbance distribution. Then, verify whether each path feature record has a usable spatial position. Delete path feature records that have no spatial position or whose spatial position falls outside the boundary of the near-arc zone. Then, map each path feature record to the corresponding spatial cell according to the position of the path center point. Statistically analyze the time deviation distribution, intensity attenuation distribution, and direction deflection distribution within each spatial cell. Output the initial distortion distribution results for S230 to call.
[0039] A spatial cell is defined as a local three-dimensional mesh cell obtained by discretizing the near-arc region under a unified reference coordinate system. Its function is to convert discrete path characteristics into a spatial parameter distribution that can be solved continuously.
[0040] In this embodiment, the side length of a single spatial unit can be 5 mm. The spatial unit into which the path center point falls determines the spatial unit to which the path feature record is assigned. When a path spans multiple spatial units, the spatial unit containing the path center point is used as the primary mapping spatial unit to avoid the same path being counted repeatedly in multiple spatial units. The three types of feature statistics within each spatial unit are generated according to the following formula: in, For the first Characteristic statistical values of each spatial unit; To map to the Number of path feature records per spatial unit; For the first The path credibility weight of each path feature record; For the first Each path feature record corresponds to a single feature value, which is any one of arrival time deviation, propagation intensity attenuation, or direction deflection. The path confidence weight can be jointly determined by propagation intensity integrity, reference path matching integrity, and direction vector stability. For example, if a path has a confidence weight of 0.8, its contribution to the statistical value of its spatial unit is greater than that of a path with a confidence weight of 0.5.
[0041] The final output of the initial distortion distribution results includes at least the spatial cell identifier, spatial cell location, cell time deviation statistics, cell propagation intensity attenuation statistics, cell orientation deflection statistics, and cell confidence level.
[0042] S230: Read the initial distortion distribution results and effective path feature set. First, verify whether each spatial unit has the unit time deviation statistics, unit propagation intensity attenuation statistics, and unit direction deflection statistics. Delete any spatial unit that is missing any statistical value or whose unit confidence is lower than the preset threshold. Then, use the combination of statistical values of the remaining spatial units as the initial particle position. Use the particle swarm algorithm to iteratively update the distortion parameters of each spatial unit in the near-arc region and output the target propagation distortion field for S310 and S410 to call.
[0043] The Particle Swarm Optimization (PSO) algorithm is used to search for the combination of spatial cell parameters that minimizes the combined residual between the measured path features and the model-predicted features. Each particle corresponds to the set of distortion parameters for all spatial cells at the current sampling time. The distortion parameters for each spatial cell include at least temporal distortion, intensity distortion, and orientation distortion parameters. The combined residual objective function is constructed as follows: in, The objective function is the composite residual. The total number of valid path feature records; Weights for arrival time deviation residuals; The weight of the residual for propagation intensity attenuation; The weight of the residual for directional deflection; , , The first The measured feature values of each path feature record; , , These are the model predicted feature values under the current particle's corresponding combination of distortion parameters.
[0044] When the particle swarm optimization algorithm reaches the preset maximum number of iterations, or when the decrease in the comprehensive residual after three consecutive iterations is less than the preset convergence threshold, it stops iterating and outputs the parameter combination corresponding to the particle that minimizes the comprehensive residual as the target propagation distortion field.
[0045] The final output target propagation distortion field includes at least the spatial cell identifier, spatial cell position, time distortion parameter, intensity distortion parameter, orientation distortion parameter, and corresponding sampling time. This output is directly used in subsequent steps to establish a distortion compensation model and correct the welding torch position, workpiece feature position, and weld position.
[0046] S3 specifically includes: S310. Read the spatial unit identifier, spatial unit position, time distortion parameters, intensity distortion parameters, and orientation distortion parameters in the target propagation distortion field. Read the absolute reference coordinates of the workstation, the nominal reference position of the workpiece, the installation pose of each sensor, and the historical calibration parameters in the initial spatiotemporal reference data. Read the actual pose of the robot end effector corresponding to the current sampling time in the near-arc multipath observation data. First, perform consistency verification on the task identifier, workstation number, welding torch configuration, and sensor configuration. Delete historical calibration parameters that are inconsistent in task identifier, workstation number, welding torch configuration, sensor configuration, or whose update time exceeds the preset update cycle. Only retain the valid historical calibration parameters that are consistent with the current welding task.
[0047] The purpose of setting effective historical calibration parameters is to provide prior geometric constraints for spatial cell position registration at the current sampling time, and to avoid errors in subsequent distortion compensation direction caused by calling historical data from other workstations or other welding gun configurations.
[0048] After verification, the spatial unit positions in the target propagation distortion field are transformed from the local coordinate system in the near-arc region to the unified reference coordinate system. The unified reference coordinate system is defined as the current welding task coordinate system established with the absolute reference coordinates of the workstation as the workstation positioning reference and the nominal reference position of the workpiece as the workpiece alignment reference. The transformation relationship is performed according to the following formula: in, For the first The position vector of each spatial unit in a unified reference coordinate system; For the first The position vector of a spatial unit in the local coordinate system of the near-arc region; This is the coordinate rotation matrix corresponding to the current workstation; This is the coordinate translation vector corresponding to the current workstation.
[0049] The purpose of the above conversion is to unify the local propagation distortion distribution obtained in the previous stage into the real space of the current workstation, and then establish a correspondence with the actual pose of the robot end effector, the nominal reference position of the workpiece, and the effective historical calibration parameters.
[0050] Subsequently, the position of each spatial unit in the unified reference coordinate system is spatiotemporally aligned with the actual pose of the robot's end effector at the current sampling time. Spatial units that do not overlap with the current working area of the welding torch are deleted, while spatial units that have a real impact on the current observation process are retained. The distortion field registration baseline data is then output for the S320 to use. The distortion field registration baseline data includes at least the spatial unit identifier, the spatial unit position in the unified reference coordinate system, the corresponding sampling time, the actual pose of the robot's end effector, the nominal reference position of the workpiece, and valid historical calibration parameters.
[0051] For example, if the current workstation number is 03 and the update time of the valid historical calibration parameter is no more than 20 sampling cycles from the current sampling time, the historical calibration parameter can be retained; if it exceeds this range, it will be deleted to ensure that the subsequent registration base data only reflects the valid status of the current workstation.
[0052] S320: Read the distortion field registration base data, and read the sensor installation pose and observation direction information in the near-arc region multipath observation dataset from the initial spatiotemporal reference data. First, determine the observation source position of each sensor in the unified reference coordinate system, and then use the line path between the observation source position and the corresponding observation object as the propagation path. Perform spatial unit crossing judgment on each propagation path. The observation objects include welding gun position observation objects, workpiece feature position observation objects, and weld position observation objects. The purpose of setting these three types of observation objects is to correspond one-to-one with the welding gun position, workpiece feature position, and weld position that need to be corrected in S410.
[0053] Records with missing observation source locations, missing observation directions, inability to establish propagation paths, or no intersection with any spatial units are deleted, retaining only path records that can be used to construct propagation influence chains. Subsequently, the spatial units traversed by the propagation path are ordered according to the order from the observation source location to the observed object along the propagation direction, forming a spatial unit action order. Then, the intersection length of the propagation path within each spatial unit is calculated, and the action weight of each spatial unit is determined according to the proportion of the intersection length to the sum of all intersection lengths. The calculation relationship is performed according to the following formula: in, For the first The weight of each spatial unit on the current propagation path; For the current propagation path in the th Intersection length within a spatial unit; The total number of spatial units traversed by the current propagation path. Let be the intersection length of the current propagation path in the j-th spatial unit.
[0054] The purpose of setting the action weight is to quantify the degree of disturbance of the propagation path in different spatial units into a comparable proportional value, so that coordinate corrections can be generated according to the same standard in the future.
[0055] For example, if a propagation path passes through three spatial units in sequence, with intersection lengths of 2 mm, 3 mm, and 5 mm, the corresponding action weights are 0.2, 0.3, and 0.5, respectively.
[0056] After calculating the action order and action weights, the propagation path identifier, observation source identifier, observation object identifier, sequence of intersecting spatial unit identifiers, action order of each spatial unit, action weight of each spatial unit, and the corresponding temporal distortion parameters, intensity distortion parameters, and directional distortion parameters of the spatial units are merged to form an observation distortion correspondence set for use by S330. Records where the sum of action weights is not equal to 1 are normalized; records with missing spatial units after normalization are deleted to ensure the integrity of the input relationships for subsequent compensation models.
[0057] S330. Read the set of observed distortion correspondences, and read the propagation path identifier, intersecting spatial unit identifier sequence, action order of each spatial unit, action weight of each spatial unit, and time distortion parameters, intensity distortion parameters, and direction distortion parameters corresponding to each spatial unit one by one. First, verify whether the relationship record has complete spatial unit parameters and complete action weights. Delete any relationship record in which any spatial unit is missing distortion parameters, any action weight is missing, or the propagation path identifier is missing, and retain the relationship records that can be directly used in the compensation calculation.
[0058] Subsequently, a spatial unit coordinate correction vector is generated for each intersecting spatial unit. The spatial unit coordinate correction vector is defined as the coordinate correction amount calculated by the time distortion parameter, intensity distortion parameter, orientation distortion parameter and corresponding sensor response coefficient of the spatial unit. Among them, the time distortion parameter is first converted into a distance correction amount along the propagation path direction, the intensity distortion parameter is converted into an attenuation compensation amount along the observation axis, and the orientation distortion parameter is converted into an angle correction amount relative to the observation direction. Then, the three are projected into a unified reference coordinate system to form the coordinate correction vector corresponding to the spatial unit.
[0059] Specifically, the coordinate correction vector for each spatial unit is calculated according to the following formula: in, This is the coordinate correction vector for the q-th spatial unit; These are the time response coefficient, intensity response coefficient, and orientation response coefficient of the current sensor, respectively. These are the temporal distortion parameter, intensity distortion parameter, and orientation distortion parameter of the q-th spatial element, respectively. To unify the unit direction vector along the observation axis in the reference coordinate system; To unify the unit normal vector perpendicular to the observation axis in the reference coordinate system.
[0060] The purpose of setting spatial unit coordinate correction vectors is to convert the propagation distortion parameters obtained in the previous stage into geometric quantities that can be directly used for coordinate reverse correction, avoiding the time, intensity, and direction parameters remaining in an intermediate state that cannot be directly compensated. After generating the coordinate correction vectors for each spatial unit, they are accumulated step by step from the observation source to the observation object according to the order of spatial unit action, and then weighted and combined according to the action weight of each spatial unit to establish a distortion compensation model for the transformation from observed coordinates to true spatial coordinates. The transformation relationship is executed according to the following formula: in, For real space coordinates; For the observed coordinates; For the first The weight of each spatial unit on the current propagation path; For the first The coordinate correction vector of each spatial unit; This represents the total number of spatial units participating in the compensation. The purpose of stepwise accumulation is to preserve the order in which disturbances occur along the propagation path, while the purpose of weighted combination is to preserve the differences in the degree of influence of different spatial units on the same observation coordinate.
[0061] The final output is a distortion compensation model for use by S410 and S510. The distortion compensation model includes at least the model identifier, applicable sampling time, applicable observation object type, observation distortion correspondence set, spatial unit coordinate correction vector generation rules, and observation coordinate to real spatial coordinate conversion rules.
[0062] S4 specifically includes: S410: Read the distortion compensation model, the target propagation distortion field, and the welding torch position observation, workpiece feature position observation, and weld position observation output by each sensor. First, verify whether each observation has an observation source identifier, observation object identifier, observation time, and observation coordinates under a unified reference coordinate system. Delete observations that are missing observation source identifiers, observation object identifiers, observation times, or observation coordinates, or that do not match the applicable sampling time of the distortion compensation model. Output the observation records to be compensated.
[0063] The purpose of setting up this verification step is to ensure that the observation records entering the compensation process and the distortion compensation model belong to the same sampling time and the same coordinate system, so as to avoid the mixing of observations from different time slices or different coordinate systems into the same correction.
[0064] Subsequently, according to the observation source identifier, the set of observation distortion correspondences and sensor response coefficients corresponding to that observation source are read. The observation records to be compensated are then input into the distortion compensation model one by one, and the corrected position corresponding to each observation record is calculated. The single-source correction position is calculated according to the following formula: in, For the object of observation At the observation source The corrected position vector below; For the object of observation At the observation source The observed position vector below; For the object of observation At the observation source The number of spatial units participating in the compensation; For the first The weight of the effect of each spatial unit on the observation location; For the first The coordinate correction vector of each spatial cell for the observation location.
[0065] The above calculation means that the coordinate offsets caused by each spatial unit traversed by the same propagation path are superimposed onto the original observation position according to their respective weights, thus obtaining the corrected position after eliminating propagation distortion. After completing the coordinate compensation, the compensation residual and the correction confidence are calculated. The compensation residual is defined as the remaining deviation between the corrected position and the original observation path constraint after being substituted back into the distortion compensation model. The correction confidence is defined as the degree of confidence jointly determined by the compensation residual, the corresponding path confidence, and the observation source stability.
[0066] Records whose corrected position exceeds the workpiece boundary, the welding torch operation boundary, or the compensation residual exceeds the preset upper limit will be deleted. For example, if the corrected position of a weld position observation record falls more than 3 mm outside the allowable welding boundary defined by the current workpiece nominal reference position, the record will be deleted directly and will not be included in the subsequent fusion.
[0067] The final output is a set of corrected location records for use by S420. The set of corrected location records includes at least the observation source identifier, the observation object identifier, the location before correction, the location after correction, the corresponding sampling time, the compensation residual, and the correction confidence level.
[0068] S420. Read the set of calibration position records, group the records according to a unified reference coordinate system and corresponding sampling time, and group the multi-source calibration positions corresponding to the same observation object at the same sampling time as a group of candidate fusion records. First, verify whether the number of observation sources in each group of candidate fusion records is not less than 2, whether each calibration position is a valid position, and whether each record carries calibration confidence. Delete candidate fusion records with less than 2 observation sources, invalid calibration positions, or missing calibration confidence, and output the group of records to be fused. The purpose of this verification step is to ensure that the unified calibration position set is supported by multiple valid observation sources, rather than being directly replaced by a single source.
[0069] Subsequently, a fusion weight was calculated for each group of records to be fused. The fusion weight was determined by the correction confidence, observation source stability, and historical bias continuity. Observation source stability characterizes the stability of output fluctuations from the same observation source across multiple sampling periods, while historical bias continuity characterizes whether the current corrected position is continuous relative to the predicted position under historical calibration parameter constraints. The unified corrected position was calculated using the following formula: in, For the object of observation The unified correction position vector; For the object of observation The number of effective observation sources; For the object of observation At the observation source The fusion weights below; For the object of observation At the observation source The corrected position vector.
[0070] The purpose of the above calculation is to give greater weight to the unified correction position by using correction positions that have higher reliability, higher stability, and are more consistent with historical deviations. After the unified correction position calculation is completed, conflict determination is performed between the multi-source correction positions of the same observation object. Conflict determination uses the Euclidean distance between the pairwise corrected positions in the unified reference coordinate system. When the Euclidean distance exceeds the conflict threshold corresponding to the observation object, the difference information is retained and marked as a high-conflict record.
[0071] For example, a 1 mm conflict threshold can be used for observing the welding torch position, a 1.5 mm conflict threshold can be used for observing the workpiece feature position, and a 1.2 mm conflict threshold can be used for observing the weld position. The final output includes a unified corrected position set and a high-conflict record set, which are available for use by S430; the unified corrected position set is also available for use by S510; the high-conflict record set includes at least the observation object identifier, conflict observation source identifier group, position difference, conflict time, and conflict level.
[0072] S430: Read the unified calibration position set, high-conflict record set, and the robot end-effector nominal pose, workpiece nominal reference position, and historical calibration parameters from the initial spatiotemporal reference data. First, group the unified calibration position set according to the observation object type to form welding torch position group, workpiece feature position group, and weld position group. Delete records that lack observation object type identification, unified calibration position, or historical calibration parameter mapping relationship, and output the constraint record to be solved. The purpose of this grouping step is to reverse-calculate the offsets corresponding to different physical objects separately, avoiding confusion between the welding torch tool center point offset, sensor extrinsic parameter offset, and robot end-effector pose mapping offset.
[0073] Subsequently, the offset of the welding torch tool center point is calculated by inversely using the difference between the unified correction position in the welding torch position group and the nominal pose of the robot end effector; the offset of the sensor extrinsic parameters is calculated by inversely using the difference between the unified correction position in the workpiece feature position group and the weld position group and the observed coordinates of each sensor; and the offset of the robot end effector pose mapping is calculated by inversely using the systematic deviation of all unified correction positions relative to the robot control pose. These three types of offsets together form the parameter offset set. The parameter offset set is solved according to the following formula: in, To calibrate the objective function of the constraints; The parameter offset set is to be determined. The parameter offset set includes at least the welding gun tool center point offset, sensor extrinsic parameter offset, and robot end-effector pose mapping offset. To standardize the number of observation objects in the calibration location set; For the object of observation The constraint credibility coefficient; For the object of observation The unified correction position vector; For the parameter offset set Observed object under action The predicted position vector, which can be obtained by combining the parameter offset set The nominal pose of the robot's end effector is substituted into a pre-established forward kinematics model of the robot to obtain the kinematics.
[0074] The constraint confidence coefficient is used to adjust the degree of influence of different observation objects on the parameter solution. When the high conflict record set shows that a certain observation object has a high conflict record at the current sampling time, the constraint confidence coefficient corresponding to the observation object is reduced. For example, the original confidence coefficient of 1 is reduced to 0.4 to weaken the bias effect of conflict records on the parameter offset solution.
[0075] After solving the objective function, the calibration constraint set is output for S510 to call. The calibration constraint set includes at least the offset of the welding gun tool center point, the offset of the sensor extrinsic parameters, the offset of the robot end pose mapping, and the constraint reliability at the corresponding sampling time.
[0076] S5 specifically includes: S510: Read the calibration constraint set, distortion compensation model, unified correction position set, and historical calibration parameters from the initial spatiotemporal reference data. First, split the calibration constraint set according to parameter type to form welding gun tool center point offset constraints, sensor extrinsic parameter offset constraints, and robot end-effector pose mapping offset constraints. Then, verify whether each constraint record simultaneously has parameter type identifier, parameter offset, corresponding sampling time, and constraint confidence. Delete constraint records with missing parameter type identifiers, missing parameter offsets, sampling times inconsistent with the applicable sampling time of the distortion compensation model, or missing constraint confidence, retaining only valid constraint records. Based on the observation object type corresponding to the unified correction position set and the compensation channel identifier in the distortion compensation model, perform parameter type mapping on the valid constraint records to form updated parameter items corresponding to the welding gun tool center point parameters, each sensor extrinsic parameter, and the robot end-effector pose mapping relationship, respectively.
[0077] Subsequently, using historical calibration parameters as the update benchmark, parameter offsets in the calibration constraint set as the increment term, and constraint reliability as the update coefficient, incremental updates are performed on the welding torch tool center point parameters, the extrinsic parameters of each sensor, and the robot end-effector pose mapping relationship to generate an updated calibration parameter set. The update relationship is executed according to the following formula: in, The updated parameter value; The corresponding parameter value in the historical calibration parameters; This refers to the parameter offsets corresponding to the calibration constraint set; These are the update coefficients corresponding to the constraint confidence levels. The update coefficients control the magnitude of parameter effectiveness, ensuring that parameters with high confidence levels undergo large updates, while parameters with low confidence levels undergo small updates.
[0078] For example, if the extrinsic parameter offset of a sensor is 0.6 mm and the update factor is 0.5, then only 0.3 mm will be updated in this cycle. After the update is completed, the updated calibration parameter set is output for the S520 to call; the updated calibration parameter set includes at least the parameter type identifier, the updated parameter value, the corresponding sampling time, the update factor, and the parameter status identifier.
[0079] S520: Read the updated calibration parameter set, distortion compensation model, and unified correction position set. First, verify whether each parameter in the updated calibration parameter set can establish a mapping relationship with the corresponding compensation channel in the distortion compensation model. Parameters that cannot establish a mapping relationship, have mismatched parameter types, or whose updated parameter values exceed the preset physical range are deleted or suspended, retaining only parameters that can be substituted back. The preset physical range is used to limit parameter updates from exceeding the allowable range of the equipment structure. For example, the update amount of the welding torch tool center point parameter in a single sampling period shall not exceed 2 mm, and the update amount of the single-axis angle in the robot end pose mapping relationship shall not exceed 1 degree.
[0080] After verification, the updated calibration parameter set will be substituted back into the distortion compensation model to regenerate the predicted position set. The predicted position set is defined as the position result obtained by recalculating the welding gun position observation object, the workpiece feature position observation object, and the weld position observation object under the current updated calibration parameter set.
[0081] Subsequently, the predicted location set and the unified correction location set are paired one by one according to the observation object identifier and sampling time, and the residual deviation of each observation object is calculated. The calculation relationship is performed according to the following formula: in, For the object of observation The remaining deviation; For the object of observation The unified correction position vector; To substitute the updated calibration parameter set back into the distortion compensation model to obtain the observed objects The predicted position vector.
[0082] Verification thresholds are set for the welding torch position observation object, the workpiece feature position observation object, and the weld position observation object. For example, the verification threshold for the welding torch position can be 0.8 mm, the verification threshold for the workpiece feature position can be 1.0 mm, and the verification threshold for the weld position can be 0.9 mm. When the remaining deviation of the corresponding observation object is not greater than its verification threshold, the relevant updated calibration parameter is marked as a passed verification parameter; when the remaining deviation of the corresponding observation object is greater than its verification threshold, the relevant updated calibration parameter is marked as a parameter to be corrected, and it is prohibited from entering the current control effective set. The parameter to be corrected is not directly deleted, but is written into the expected correction list for the next sampling cycle along with the remaining deviation value, failure type identifier, and corresponding sampling time.
[0083] The final output includes the set of calibration parameters that have passed the verification and the parameter verification results, which can be called by S530. The parameter verification results include at least the parameter type identifier, the corresponding observation object type, the remaining deviation value, the verification threshold, the verification conclusion, and the verification confidence level.
[0084] S530: Read the calibration parameter set that has passed verification, the parameter verification results, and the historical calibration parameters in the initial spatiotemporal reference data. First, verify whether each parameter in the calibration parameter set that has passed verification has a parameter type identifier, parameter value, corresponding sampling time, and parameter status identifier. Delete records with missing parameter status identifiers, missing parameter values, or missing corresponding sampling times. At the same time, read the verification confidence level in the parameter verification results. Keep the historical calibration parameters unchanged for parameters that have failed verification. Generate position self-calibration results for parameters that have passed verification and write them into the robot control system according to parameter type.
[0085] The position self-calibration result is defined as the set of calibration results that can be directly entered into the robot control effective set at the current sampling time. It includes at least the result identifier, applicable sampling time, welding torch tool center point parameters, sensor external parameters, robot end-effector pose mapping relationship, calibration reliability of this cycle, and parameter status identifier. Among them, the parameter status identifier is used to distinguish whether the parameter belongs to "updated after verification in this cycle" or "using the parameters verified in the previous cycle".
[0086] To avoid amplifying the overall result error by a small number of low-quality observations, the calibration reliability for this period is calculated based on all observations involved in the verification. The calculation relationship is as follows: in, To determine the reliability of this period; The number of observation objects participating in the verification; For the object of observation The verification weight; For the object of observation The credibility of single-object verification.
[0087] After completing the calibration reliability calculation for this cycle, the verified welding torch tool center point parameters, sensor extrinsic parameters, and robot end-effector pose mapping relationship will be written into the current control active set of the robot control system, and the position self-calibration results for this cycle will be updated to the historical calibration parameter library according to the result identifier and applicable sampling time.
[0088] Therefore, the position self-calibration result is used for subsequent arc welding trajectory correction and weld repositioning, and in the next sampling cycle, it is used as the historical calibration parameters read by S110 to enter a new round of processing.
[0089] In this embodiment, k (or All relevant subscripts are positive integers, representing the sequence number of the corresponding record, unit, or object.
[0090] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.
[0091] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.
[0092] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0093] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0094] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.
[0095] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0096] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.
[0097] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0098] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0099] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for multi-sensor position self-calibration of an embodied robot for arc welding operations, characterized in that, Includes the following steps: S1. Read the Beidou positioning output, robot end nominal pose, workpiece nominal reference position, sensor installation pose, historical calibration parameters and arc-free reference path set corresponding to the welding task, collect and complete the time alignment and spatial alignment of the original multipath measurement records, and obtain the near-arc area multipath observation dataset. S2. Based on the near-arc region multipath observation dataset and the arc-free reference path set, the arrival time deviation, propagation intensity attenuation and direction deflection are extracted to generate an effective path feature set, and the initial distortion distribution results are formed according to spatial units. S3. Using the initial distortion distribution results as the initial solution, the particle swarm optimization algorithm is used to solve the target propagation distortion field, and an observation distortion correspondence set and distortion compensation model are established. S4. Based on the distortion compensation model, reverse correction and fusion are performed on the welding gun position, workpiece feature position and weld position to obtain a unified correction position set and a high conflict record set, and a calibration constraint set is generated. S5. Update the center point parameters of the welding torch tool, sensor extrinsic parameters and robot end pose mapping relationship according to the calibration constraint set, and output the position self-calibration result after back-substitution verification.
2. The method for multi-sensor position self-calibration of a unibody robot for arc welding operations according to claim 1, characterized in that, S1 specifically includes: Read the Beidou positioning output, robot end nominal pose, workpiece nominal reference position, sensor installation pose, historical calibration parameters and arc-free reference path set corresponding to the welding task to form initial spatiotemporal reference data; Based on the initial spatiotemporal reference data, multiple sensors are controlled to transmit, receive, or sample synchronously under a unified time reference, generating a set of original multipath measurement records containing propagation path identifiers and corrected acquisition times.
3. The method for multi-sensor position self-calibration of a unibody robot for arc welding operations according to claim 2, characterized in that, Also includes: The original multipath measurement record set is combined with the arc-free reference path set to perform time alignment, spatial alignment and observation unit merging. Observation units without reference paths or with insufficient number of effective propagation paths are deleted, and the near-arc region multipath observation dataset is output.
4. The method for multi-sensor position self-calibration of a unibody robot for arc welding operations according to claim 1, characterized in that, S2 specifically includes: Read the near-arc region multipath observation dataset and the arc-free reference path set, perform same-path identifier pairing verification for each propagation path, delete propagation path records that cannot be matched, cannot construct direction vectors, or have missing fields, and calculate arrival time deviation, propagation intensity attenuation and direction deflection to generate an effective path feature set; Based on the effective path feature set and the aligned spatial position of each propagation path, the features of each path are mapped to the corresponding spatial unit. The time deviation, propagation intensity attenuation and directional deflection of each spatial unit are statistically analyzed to form the initial distortion distribution result.
5. A multi-sensor position self-calibration method for a unibody robot for arc welding operations according to claim 4, characterized in that, Also includes: Using the initial distortion distribution as the initial solution, the time distortion parameters, intensity distortion parameters, and orientation distortion parameters of each spatial unit are iteratively updated using the particle swarm optimization algorithm to obtain the target propagation distortion field.
6. The method for multi-sensor position self-calibration of a unibody robot for arc welding operations according to claim 1, characterized in that, S3 specifically includes: Read the target propagation distortion field, initial spatiotemporal reference data, absolute reference coordinates of the workstation, nominal reference position of the workpiece, and actual pose of the robot end effector. Perform coordinate transformation, task consistency verification, and deletion of invalid spatial units on the spatial unit positions to form the basic data for distortion field registration. Based on the registration data of the distortion field, the spatial units traversed by the propagation path, the order of action, and the weight of action are determined, and a set of observation distortion correspondences between the observation source, the observation object, and the distortion parameters of the spatial unit is established.
7. A multi-sensor position self-calibration method for a unibody robot for arc welding operations according to claim 6, characterized in that, Also includes: Based on the set of observed distortion correspondences, coordinate correction vectors for each spatial unit are generated, and a distortion compensation model for transforming observed coordinates into true spatial coordinates is established according to the order of action and the weight of action.
8. The method for multi-sensor position self-calibration of a unibody robot for arc welding operations according to claim 1, characterized in that, S4 specifically includes: Read the distortion compensation model, perform compensation correction on the welding torch position observation value, workpiece feature position observation value and weld position observation value output by each sensor, delete the records that exceed the workpiece boundary, welding torch operation boundary or compensation residual limit after correction, and generate a set of corrected position records. The set of corrected location records is grouped and fused according to the observation object and sampling time to obtain a unified set of corrected locations, and the set of high-conflict records is marked according to the distance difference between the multi-source corrected locations. Based on a unified set of corrected positions, a set of high-conflict records, the nominal pose of the robot's end effector, and historical calibration parameters.
9. A multi-sensor position self-calibration method for a unibody robot for arc welding operations according to claim 8, characterized in that, Also includes: Solve for the offset of the welding gun tool center point, the offset of the sensor extrinsic parameters, and the offset of the robot end-effector pose mapping to form a calibration constraint set.
10. A multi-sensor position self-calibration method for a unibody robot for arc welding operations according to claim 1, characterized in that, S5 specifically includes: Read the calibration constraint set, distortion compensation model, unified correction position set and historical calibration parameters, split the parameter offsets according to parameter type, and update the welding gun tool center point parameters, sensor extrinsic parameters and robot end pose mapping relationship according to constraint confidence, and generate updated calibration parameter set; The updated calibration parameter set is substituted back into the distortion compensation model to form a predicted location set. This set is then compared with the unified correction location set to obtain the remaining deviation and parameter verification results. Finally, the calibration parameter set that passes the verification is selected. The position self-calibration result is generated based on the verified set of calibration parameters and the parameter verification results, and written into the robot control system and the historical calibration parameter library.