Multi-axis collaborative welding robot dynamic path planning method and system

Abstract

This application relates to the field of welding robot technology, specifically to a dynamic path planning method and system for a multi-axis collaborative welding robot. The method includes: deploying a displacement sensor array to synchronously acquire three-dimensional displacement data; constructing a global thermal deformation field for the workpiece using the path distance of the structural skeleton as an interpolation weight benchmark; performing correction vector calculation based on the global thermal deformation field of the workpiece at the same timestamp, calculating the local gradient magnitude of the deformation field at each path point; determining an adaptive safety margin for the deformation gradient based on the local gradient magnitude of the deformation field, and performing collaborative conflict detection; and using a rolling temporal planning framework for periodic iterative updates, concentrating computational power on active thermal deformation regions after the correction vector converges. This application effectively avoids path conflicts caused by directional deviations of correction vectors between adjacent robots in a shared workspace.

Description

Technical Field

[0001] This application relates to the field of welding robot technology, specifically to a dynamic path planning method and system for a multi-axis collaborative welding robot. Background Technology

[0002] In the field of sheet metal welding, products such as smart express cabinet bodies, home appliance shells, and automotive stamping parts generally have multi-faceted intersecting welds and complex spatial structures. A single robot cannot meet the production cycle requirements, so the collaborative operation of multiple welding robots has become the mainstream deployment method. In such production scenarios, by having multiple robots simultaneously weld different areas of the same workpiece, the welding cycle for a single part can be significantly reduced.

[0003] To address the inherent dimensional tolerance issues of sheet metal stamping parts, existing technologies generally integrate laser displacement sensors at the front end of each robot's welding torch. By collecting local contour data of the weld seam in real time, online deviation compensation is performed on the pre-programmed path, i.e., weld seam tracking correction. This solution is relatively mature in single-robot operation scenarios and can effectively cope with static position errors caused by workpiece clamping deviations.

[0004] However, because welding heat input continuously accumulates inside the workpiece and induces nonlinear thermal deformation, the actual geometry of the workpiece is in a state of dynamic drift throughout the welding process. Simultaneously, due to the global spatial correlation of the workpiece's thermal deformation field, thermal expansion in one region inevitably propagates to adjacent regions through the structural framework. This means that the deformation state perceived by each robot based on its local sensors is essentially only a local slice of the same global deformation field at different spatial locations and sampling times. When each robot uses these spatiotemporally inconsistent local slices as independent criteria for path correction, the path correction vectors of adjacent robots in the shared workspace will exhibit directional deviations. If the correction direction of one robot happens to point to the corrected trajectory of another robot, it will cause path conflicts between the two in the deformed actual space, or the weld seam offset will exceed tolerances due to the mutual cancellation of correction directions. This problem is particularly prominent in situations where multiple robots weld simultaneously with high thermal deformation rates. Summary of the Invention

[0005] To address the existing technical problem in multi-robot collaborative welding where each robot independently corrects its path based on local deformation perception, resulting in directional conflicts in the path correction vectors of adjacent robots within the shared workspace due to the inability to share global thermal deformation field information, this application provides a dynamic path planning method and system for multi-axis collaborative welding robots.

[0006] In a first aspect, this application provides a dynamic path planning method for a multi-axis collaborative welding robot, comprising: arranging a displacement sensor array based on the topology of the sheet metal skeleton of the workpiece, synchronously acquiring three-dimensional displacement, and constructing a global thermal deformation field of the workpiece using the path distance of the skeleton as an interpolation weight benchmark; performing a unified correction vector calculation on the pre-programmed paths of all robots based on the global thermal deformation field of the workpiece at the same timestamp, obtaining the corrected actual target coordinates, and calculating the local gradient magnitude of the deformation field at each path point; determining an adaptive safety margin for the deformation gradient based on the local gradient magnitude of the deformation field and the basic safety radius, and performing collaborative conflict detection on the corrected paths of all robots based on the adaptive safety margin of the deformation gradient; using a rolling temporal planning framework to periodically iteratively update the global thermal deformation field of the workpiece and the path correction, and concentrating computing power on the active thermal deformation region after the correction vector converges.

[0007] By introducing the structural skeleton path distance to replace the Euclidean distance for deformation field reconstruction, the mismatch between the physical characteristics of thermal deformation transmission along the structure and the traditional interpolation assumptions is solved, making the deformation field reconstruction results more realistically reflect the actual heat conduction path of the sheet metal parts. Furthermore, the adaptive safety margin of the deformation gradient is incorporated into the conflict detection criterion, ensuring that cooperative collision avoidance remains effective in the dynamic environment of thermal deformation. At the same time, a rolling temporal programming framework is used to periodically update the deformation field and path correction, and after the correction vector converges, the computing power is concentrated on the active thermal deformation region, thus taking into account both the real-time performance of path correction and the reasonable allocation of computing resources.

[0008] Preferably, the workpiece-based sheet metal skeleton topology arrangement of displacement sensor array, synchronously acquiring three-dimensional displacement, and constructing a global thermal deformation field of the workpiece using the skeleton path distance as the interpolation weight benchmark, includes: arranging displacement sensor arrays at each major structural node, extracting the maximum deformation gradient value between adjacent structural nodes; determining the upper limit of the sensor node spacing based on the maximum deformation gradient value; synchronously acquiring the three-dimensional displacement at the current moment through each sensor node to form a globally synchronized deformation snapshot; and constructing the global thermal deformation field of the workpiece using the structural path weighted inverse distance interpolation method based on the synchronously acquired sparse node displacement data. Preferably, the construction of the global thermal deformation field of the workpiece using the structural path weighted inverse distance interpolation method based on the synchronously acquired sparse node displacement data includes: abstracting the workpiece skeleton as a directed graph, using the shortest path algorithm to solve the shortest skeleton path distance between all node pairs offline; obtaining the shortest transmission path length along the skeleton from the sensor node to the point to be solved; and calculating the reconstructed displacement vector at the point to be solved based on the three-dimensional displacement and the shortest transmission path length.

[0009] This enables the deformation field reconstruction results to accurately reflect the physical path of thermal deformation transmission along the sheet metal skeleton, effectively avoiding reconstruction distortion caused by underestimating the transmission impedance due to Euclidean distance when crossing bending edges or hollow areas, thus providing a self-consistent unified data base for all subsequent robot path corrections.

[0010] Preferably, the step of performing a unified correction vector calculation on all robot pre-programmed paths based on the global thermal deformation field of the workpiece at the same timestamp to obtain the corrected actual target coordinates, and calculating the local gradient magnitude of the deformation field at each path point, includes: obtaining the original design coordinates of the robot's pre-programmed path; calculating the corrected actual target coordinates based on the original design coordinates and the interpolated displacement vector at those coordinates from the reconstructed global deformation field; and using the finite difference method, calculating the degree of spatial change of the deformation field near the path point based on the reconstructed deformation field data, as the local gradient magnitude of the deformation field. Preferably, after calculating the local gradient magnitude of the deformation field at each path point, the step further includes: performing a rationality check on path points with abnormally high gradient magnitudes; if the gradient magnitude at a certain path point exceeds a preset multiple threshold of the average gradient magnitude of adjacent path points, replacing the gradient magnitude at that point with the average gradient magnitude of adjacent path points.

[0011] This prevents subsequent safety margin calculations from becoming excessively inflated due to local noise spikes, thus avoiding unnecessary compression of the workspace.

[0012] Preferably, the step of determining the adaptive safety margin of the deformation gradient based on the local gradient magnitude of the deformation field and the basic safety radius, and performing cooperative conflict detection on the corrected paths of all robots based on the adaptive safety margin of the deformation gradient, includes: obtaining the skeleton path distance from the point to the nearest sensor node; calculating the adaptive safety margin of the deformation gradient based on the basic safety radius, the skeleton path distance, and the local gradient magnitude of the deformation field; if there is a pair of path points whose spatial distance is less than the sum of the corresponding adaptive safety margins of the deformation gradient, it is determined as a potential conflict point pair, and the timing scheduling intervention is triggered. Preferably, the triggering of the timing scheduling intervention includes: applying sequential constraints to the conflicting path point pairs on the time axis; causing one robot to pause and wait before reaching the conflict area, and then continue to advance after the other robot passes through the conflict area and exits the safety margin range; after processing each conflict point pair, updating the estimated arrival time of all subsequent path points of that robot, and re-performing conflict detection until there are no more conflict point pairs in the current planning time domain.

[0013] Replacing spatial domain detour with temporal domain separation to handle conflicts ensures that cooperative collision avoidance remains effective in dynamic environments of thermal deformation, while also preventing the introduction of new path deviations due to spatial detours.

[0014] Preferably, the step of periodically iteratively updating the global thermal deformation field and path correction of the workpiece using a rolling time-domain planning framework, and concentrating computing power on the active thermal deformation region after the correction vector converges, includes: obtaining the maximum thermal deformation rate among all nodes; determining the planning period based on the maximum thermal deformation rate and the basic safety radius; triggering global synchronous sampling at the beginning of each planning period to obtain the latest deformation field snapshot, and recalculating the correction vectors and conflict detection results for a preset number of future path points. Preferably, after recalculating the correction vectors and conflict detection results for a preset number of future path points, the step further includes: monitoring the change amplitude of the correction vector of the same path point within two adjacent planning periods; when the change amplitude is lower than a preset multiple of the standard deviation of the sensor measurement noise, determining that the correction vector of the path point has converged; after determining convergence, the point will not be recalculated in subsequent planning periods.

[0015] Secondly, this application provides a dynamic path planning system for a multi-axis collaborative welding robot, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aforementioned dynamic path planning method for a multi-axis collaborative welding robot.

[0016] By adopting the above technical solution, a computer program is generated from the above-mentioned dynamic path planning method for multi-axis collaborative welding robots and stored in a memory so that it can be loaded and executed by a processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0017] This application ensures data consistency among multiple robots in a shared workspace by using unified timestamp alignment and correction vector calculation, effectively eliminating path correction direction conflicts caused by local perception differences, and improving the stability and safety of collaborative operations.

[0018] Meanwhile, the introduction of deformation gradient adaptive safety margin and rolling time-domain planning framework can flexibly respond to potential collision risks in the dynamically evolving thermal deformation environment. The timing scheduling avoids the additional deviation caused by spatial detours, and the dynamic allocation of computing power based on convergence criteria ensures the timeliness of path correction and the efficient use of computing resources, meeting the actual needs of efficient collaborative welding of complex sheet metal parts. Attached Figure Description

[0019] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1This is a flowchart illustrating a dynamic path planning method for a multi-axis collaborative welding robot according to the present invention.

[0020] Figure 2 This is a comparison chart of the convergence process of the path correction residual with the iteration cycle in the embodiment of the present invention and the effect of the prior art. Detailed Implementation

[0021] 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, not all, of the embodiments of the present invention. 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.

[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] This invention discloses a dynamic path planning method for a multi-axis cooperative welding robot, referring to... Figure 1 This includes steps S1-S4: S1. Construct the workpiece thermal deformation field based on structural path weighted interpolation.

[0024] In an optional embodiment, before the welding operation begins, displacement sensor arrays are arranged at each major structural node according to the sheet metal skeleton topology of the workpiece to form a sparse sensing network. The major structural nodes include the intersection points of bending edges, the start and end points of welds, and the center points of large-area plates. During the welding process debugging phase, the full-field thermal deformation distribution of the workpiece under rated welding current is measured, and the maximum deformation gradient value between adjacent structural nodes is extracted. This is used to characterize the displacement change per unit skeleton path length, with dimensions in mm / mm. According to... Determine the upper limit of sensor node spacing It must satisfy:

[0025] Furthermore, during the welding process, the three-dimensional displacement at the current moment is synchronously acquired through each sensor node at a fixed sampling period. To create a globally synchronized deformation snapshot, and to eliminate inconsistencies in the temporal dimension of local perception among different robots, the sampling times of all nodes must be aligned to the same timestamp. To address the issue of sampling time deviation, if the deviation between the sampling time of any node and the global timestamp exceeds one-tenth of the sampling period, the sampling data of that node in this current sampling is marked as invalid, and the data from the previous valid time is used instead, ensuring the temporal consistency of the deformation snapshot.

[0026] Next, based on the synchronously acquired sparse node displacement data, a structural path weighted inverse distance interpolation method is used to construct the global thermal deformation field of the workpiece. Because the original inverse distance weighted interpolation algorithm uses the Euclidean straight-line distance from the sensor node to the point to be solved. As the benchmark for weight calculation, its interpolation formula is:

[0027] Because thermal deformation is conducted along the sheet metal structural skeleton rather than propagating in a straight line in space, directly using Euclidean distance would overestimate the deformation correlation across bending edges or cutout areas, leading to reconstruction distortion. Therefore, the weight calculation benchmark is replaced with the shortest path distance along the workpiece structural skeleton. That is, from the sensor node To the point to be solved The improved interpolation formula is obtained by finding the shortest conduction path length along the structural skeleton:

[0028] In the above relation, Point to be solved The reconstructed displacement vector at that location For the first The measured displacement vector of each sensor node. This represents the total number of sensor nodes.

[0029] against The calculation pre-abstracts the workpiece structural skeleton into a directed graph, where nodes are structural intersections and edge weights are the actual lengths of the corresponding structural components. Dijkstra's algorithm is used offline to solve for the shortest skeleton path distance between all node pairs, and the result is stored as a distance matrix for online interpolation. It should be noted that many algorithms in this field can solve for the shortest path on a graph. Those skilled in the art can choose different shortest path algorithms based on the size and sparsity of the workpiece skeleton graph; Dijkstra's algorithm is just one possible implementation.

[0030] In this way, by using the structural skeleton path distance instead of the Euclidean distance as the interpolation weight benchmark, the deformation field reconstruction results can truly reflect the physical path of thermal deformation transmission along the sheet metal skeleton. This effectively avoids reconstruction distortion caused by the underestimation of transmission impedance due to the Euclidean distance when crossing bending edges or hollow areas, thus providing a physically self-consistent unified data base for all subsequent robot path corrections.

[0031] S2. Calculate the correction vector and gradient magnitude based on the unified deformation field.

[0032] In an optional embodiment, when obtaining the global thermal deformation field Then, a unified correction vector calculation is performed on the pre-programmed paths of all robots, assuming the first... The pre-programmed path of the robot consists of a series of path points If the structure is as follows, then the coordinates of each path point after deformation correction are:

[0033] In the above relation, For the first Taiwan Robot The original design coordinates of each path point This is the interpolated displacement vector at this coordinate, derived from the reconstructed global deformation field. These are the corrected actual target coordinates. To fundamentally eliminate the temporal inconsistency between the correction vectors of each robot and ensure that all correction vectors are spatially consistent in the same global coordinate system, the correction vectors of all robots must be based on a deformation field snapshot at the same timestamp. Perform the solution.

[0034] Furthermore, after completing the correction vector calculation, the local gradient magnitude of the deformation field is calculated for each corrected path point. This gradient magnitude is used to characterize the spatial drastic changes in the deformation field near the path point and serves as the input for subsequent dynamic safety margin calculations. The gradient magnitude is approximated using the finite difference method, with the path point as the reference point. Centered on, along , and Step size in three coordinate axes The calculation formula is:

[0035] Among them, step size The spacing between adjacent points on the pre-programmed path is taken, and the above gradient magnitude calculations are all based on the reconstructed deformation field data.

[0036] Next, the rationality of path points with abnormally high gradient amplitudes in the deformation field needs to be verified. If the gradient amplitude at a certain path point exceeds a preset multiple threshold of the average gradient amplitude of adjacent path points, it indicates that the point may be affected by local sensor noise. In this case, its gradient amplitude is replaced with the average gradient amplitude of adjacent path points to prevent excessive expansion of the subsequent safety margin calculation due to local noise spikes, thus avoiding unnecessary compression of the workspace. For example, the preset multiple threshold can be 3. For example, the value range of adjacent path points can be the path points themselves. Two path points before and after, i.e., take , , and The average gradient magnitude of the four points was used as the replacement benchmark.

[0037] In this way, by forcing all robots to uniformly calculate the correction vector based on the same timestamp of the deformation field snapshot and synchronously outputting the gradient magnitude of the deformation field at each path point, the inconsistency of multi-robot path correction is eliminated from the time dimension, and a quantitative input reflecting the distribution of spatial reconstruction uncertainty is provided for subsequent conflict detection, enabling collaborative path planning to have the ability to perceive the local complexity of the deformation field.

[0038] S3. Perform collaborative conflict detection based on deformation gradient adaptive safety margin.

[0039] In an optional embodiment, cooperative conflict detection is performed on the corrected paths of all robots in the corrected coordinate space. Since the accuracy of deformation field reconstruction decreases with increasing distance from the sensor node's skeleton path, regions far from the sensor node exhibit greater reconstruction uncertainty. If a traditional method with a fixed safety radius is used... A collision detection scheme based solely on this criterion will fail to reflect the uncertainty of this spatially uneven distribution. Therefore, an adaptive safety margin for deformation gradients is defined. :

[0040] In the above relation, The basic safety radius is the kinematic repeatability accuracy of the robot body. It is directly taken from the repeatability accuracy index in the robot's manufacturer's specifications and has the dimension of mm. For point The skeleton path distance to the nearest sensor node, in mm; Let be the magnitude of the deformation field gradient at that point, with dimensions in mm / mm.

[0041] Furthermore, for any two robots and The corrected path, if a pathpoint pair exists. The spatial distance is less than If a potential conflict point is identified, a time-series scheduling intervention is triggered. The time-series scheduling process is as follows: a sequential constraint is imposed on the conflict path point pair on the time axis, causing one robot to pause and wait before reaching the conflict area. The other robot can continue its progress after passing through the conflict area and leaving the safety margin range. This replaces spatial detour with temporal domain separation, avoiding the introduction of additional spatial path deformation.

[0042] Next, for cases where multiple conflict pairs exist within the same planning time domain for the same pair of robots, sequential scheduling is used to process the conflict pairs according to their order on their respective paths. Since processing the nearest conflict pairs first introduces a waiting delay that shifts the robot's arrival time at subsequent path points, failing to re-verify the timing relationships of subsequent conflict pairs could lead to new conflicts arising from the time shift in previously non-conflicting path points. Therefore, after processing each conflict pair, the estimated arrival times of all subsequent path points for that robot are updated, and conflict detection is re-executed until no more conflict pairs exist within the current planning time domain.

[0043] Thus, by incorporating the uncertainty of deformation field reconstruction into the conflict detection criterion in the form of gradient adaptive safety margin, and replacing spatial domain detour processing with time domain separation, cooperative collision avoidance maintains its effectiveness in the dynamic environment of thermal deformation, avoids the introduction of new path deviations due to spatial detour, and prevents excessive compression of the working space through the shrinkage characteristics of the adaptive margin.

[0044] S4. The deformation field is iteratively updated and the convergence path is corrected using a rolling time-domain framework.

[0045] In an optional embodiment, since the welding heat input continuously accumulates as the operation progresses, the workpiece thermal deformation field is in a dynamic evolution state, causing the reconstructed deformation field snapshot to gradually become invalid over time. Therefore, a rolling time-domain planning framework is used to periodically iteratively update the deformation field and path correction. Regarding the planning cycle... The value of is determined during the welding process debugging phase by measuring the thermal deformation rate of each structural node of the workpiece under the rated welding current, recording the displacement change of each node per unit time, and taking the maximum thermal deformation rate among all nodes as . The unit is mm / s, and Must meet:

[0046] That is, within a planning period, the increase in thermal deformation at any node of the workpiece does not exceed the basic safety radius. This ensures that the error of the correction vector calculated in the previous cycle is always within the compensation range of the safety margin. The specific values ​​are directly derived from the above inequalities and are recalibrated as the workpiece material or welding parameters change.

[0047] Furthermore, at the beginning of each planning cycle, a global synchronization sampling is triggered to obtain the latest deformation field snapshot, and the future deformation field is recalculated according to the aforementioned process. The correction vectors for each path point and the conflict detection results are used to send the solution results to each robot controller to overwrite the planning results of the previous cycle. Planning time domain. The selection is based on system communication latency. and the time consumed by a single planning calculation Determined, making The estimated execution time for each path point is greater than This means that the effective coverage time of the current planning result fully includes the entire delay of the sampling, calculation, and transmission chain, ensuring that the robot can continue moving without stopping until the next planning result arrives. and All data were obtained through actual testing during the system integration and debugging phase.

[0048] Subsequently, as the welding operation progresses, the thermal deformation of the areas corresponding to the completed path points of each robot tends to stabilize. After the heat source moves away, the area enters a cooling phase, causing the accuracy of the deformation field reconstruction in this region to continuously improve with the accumulation of sensor data. The calculation error of the path correction vector shows a convergence trend. By monitoring the change amplitude of the correction vector of the same path point within two adjacent planning cycles, when this change amplitude is lower than the standard deviation of the sensor measurement noise... When the preset multiple is reached, it is determined that the correction vector of the path point has converged. For example, the preset multiple can be 2. Since the change in the correction vector is already submerged in the sensor noise floor at this point, further iteration will no longer produce a substantial correction effect. The values ​​are directly obtained from the sensor's factory calibration data. After convergence is determined, the point will not be recalculated in subsequent planning cycles, thus concentrating computing power on future path points that are still in the active thermal deformation region.

[0049] Figure 2 This is a comparison chart of the convergence process of the path correction residual with the iteration cycle in the embodiment of the present invention and the effect of the prior art. It can be seen that the residual curve of the method of the present application converges much faster than the prior art, and can be stably reduced to below the engineering accuracy threshold in a shorter iteration cycle.

[0050] In this way, by incorporating deformation field updates, path correction calculations, and conflict detection into a unified periodic iterative process through a rolling time-domain planning framework, and dynamically releasing computational resources in the stable region based on the convergence criterion of sensor noise background, the system can maintain the timeliness of path correction throughout the entire welding operation cycle, while avoiding redundant calculations in the converged region, and allowing limited computing power to continuously focus on the most active frontier region of thermal deformation.

[0051] This invention also discloses a dynamic path planning system for a multi-axis collaborative welding robot, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a dynamic path planning method for a multi-axis collaborative welding robot according to the present invention.

[0052] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A multi-axis collaborative welding robot dynamic path planning method, characterized in that, include: A displacement sensor array is arranged based on the topology of the sheet metal structure skeleton of the workpiece to synchronously collect three-dimensional displacement, and the global thermal deformation field of the workpiece is constructed using the path distance of the structure skeleton as the interpolation weight benchmark. Based on the global thermal deformation field of the workpiece at the same timestamp, a unified correction vector solution is performed on the pre-programmed paths of all robots to obtain the corrected actual target coordinates, and the local gradient magnitude of the deformation field at each path point is calculated. Based on the local gradient magnitude of the deformation field and the basic safety radius, the adaptive safety margin of the deformation gradient is determined, and cooperative conflict detection is performed on the corrected paths of all robots based on the adaptive safety margin of the deformation gradient. A rolling time-domain programming framework is used to periodically update the global thermal deformation field and path correction of the workpiece, and the computing power is concentrated on the active thermal deformation region after the correction vector converges.

2. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 1, characterized in that, The workpiece-based sheet metal frame topology arrangement of the displacement sensor array synchronously acquires three-dimensional displacement, and constructs the workpiece's global thermal deformation field using the frame path distance as the interpolation weight benchmark, including: Displacement sensor arrays are deployed at each major structural node to extract the maximum deformation gradient value between adjacent structural nodes; The upper limit of the sensor node spacing is determined based on the maximum deformation gradient value; The three-dimensional displacement at the current moment is collected synchronously by each sensor node to form a globally synchronized deformation snapshot; Based on synchronously acquired sparse node displacement data, a structural path weighted inverse distance interpolation method is used to construct the global thermal deformation field of the workpiece.

3. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 2, characterized in that, The sparse node displacement data acquired synchronously is used to construct the global thermal deformation field of the workpiece using a structure path weighted inverse distance interpolation method, including: The workpiece structural skeleton is abstracted as a directed graph, and the shortest path algorithm is used to solve the shortest skeleton path distance between all node pairs offline. Obtain the shortest transmission path length along the structural skeleton from the sensor node to the point to be solved; Based on the three-dimensional displacement and the shortest transmission path length, the reconstructed displacement vector at the point to be solved is calculated.

4. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 1, characterized in that, The global thermal deformation field of the workpiece based on the same timestamp is used to perform a unified correction vector calculation on the pre-programmed paths of all robots to obtain the corrected actual target coordinates, and the local gradient magnitude of the deformation field at each path point is calculated, including: Obtain the original design coordinates of the robot's pre-programmed path; Based on the original design coordinates and the interpolated displacement vector at those coordinates obtained from the reconstructed global deformation field, the corrected actual target coordinates are calculated. Using the finite difference method, based on the reconstructed deformation field data, the degree of spatial variation of the deformation field near the path point is calculated and used as the local gradient magnitude of the deformation field.

5. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 4, characterized in that, After calculating the local gradient magnitude of the deformation field at each path point, the method further includes: For path points with abnormally high gradient magnitudes in the deformation field, a rationality check is performed. If the gradient magnitude at a certain path point exceeds a preset multiple threshold of the average gradient magnitude of adjacent path points, the gradient magnitude at that point will be replaced with the average gradient magnitude of adjacent path points.

6. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 1, characterized in that, The process of determining an adaptive safety margin for the deformation gradient based on the local gradient magnitude of the deformation field and the basic safety radius, and performing cooperative conflict detection on the corrected paths of all robots based on the adaptive safety margin for the deformation gradient, includes: Obtain the skeleton path distance from the point to the nearest sensor node; Based on the basic safety radius, the skeleton path distance, and the local gradient magnitude of the deformation field, the adaptive safety margin of the deformation gradient is calculated. If there exists a pair of path points whose spatial distance is less than the sum of the corresponding deformation gradient adaptive safety margins, it is identified as a potential conflict pair and time-series scheduling intervention is triggered.

7. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 6, characterized in that, The triggering of the timing scheduling intervention includes: Apply sequential constraints to pairs of conflicting path points on the timeline; One robot is instructed to pause and wait before reaching the conflict zone, while the other robot continues its advance only after it has passed through the conflict zone and exited the safe margin. After processing each conflict point pair, the estimated arrival times of all subsequent path points of the robot are updated, and conflict detection is re-executed until there are no more conflict point pairs in the current planning time domain.

8. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 1, characterized in that, The method of periodically iteratively updating the global thermal deformation field and path correction of the workpiece using a rolling time-domain programming framework, and concentrating computing power on the active thermal deformation region after the correction vector converges, includes: Obtain the maximum thermal deformation rate among all nodes; The planning period is determined based on the maximum thermal deformation rate and the basic safety radius; At the beginning of each planning cycle, global synchronous sampling is triggered to obtain the latest deformation field snapshot, and the correction vectors and collision detection results of a preset number of path points are recalculated.

9. The dynamic path planning method for a multi-axis cooperative welding robot according to claim 8, characterized in that, After recalculating the correction vectors and conflict detection results for a predetermined number of future path points, the method further includes: Monitor the magnitude of the change in the correction vector of the same path point within two adjacent planning periods; When the change amplitude is lower than a preset multiple of the standard deviation of the sensor measurement noise, it is determined that the correction vector of the path point has converged; Once convergence is determined, the point will not be recalculated in subsequent planning cycles.

10. A dynamic path planning system for a multi-axis collaborative welding robot, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute a computer program stored in the memory, implements the dynamic path planning method for a multi-axis collaborative welding robot as described in any one of claims 1 to 9.

Images