Dynamic calibration of large-scale space and hand-eye error compensation method for isomorphic dual-robot arms

Abstract

The present application relates to the technical field of industrial robot cooperative control, and more particularly to a heterogeneous dual-robot large-scale space dynamic calibration and hand-eye error compensation method, comprising the following steps: acquiring real-time poses of two heterogeneous robots, workpiece space states and visual observation data, generating cooperative space state data through time synchronization and space alignment; relying on the data and dynamic constraints of the robots on the workpiece, constructing and real-time updating a workpiece-level dynamic process reference constraint system; combining the robot poses and the reference constraint system to establish a reference transmission chain and reconstruct the cooperative space mapping relationship; updating the dynamic hand-eye mapping relationship according to the mapping relationship, and generating and adaptively correcting the machining trajectory of the second heterogeneous robot. The present application fuses the poses of the two heterogeneous robots and the visual observation data, constructs a dynamic workpiece process reference constraint system, realizes cross-robot space mapping and adaptive updating of the hand-eye, and solves the problems of mapping mismatch and error accumulation in traditional static calibration.

Description

Technical Field

[0001] This invention relates to the field of collaborative control technology for industrial robots, and in particular to a method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms. Background Technology

[0002] With the development of intelligent manufacturing and flexible production technologies, heterogeneous dual-arm collaborative operations are increasingly being applied in scenarios such as complex assembly, precision machining, and large-scale spatial operations. In these applications, different types of robotic arms typically undertake tasks such as handling, positioning, or processing, and acquire workpiece status information through a vision system to achieve collaborative control of multiple devices. Simultaneously, to ensure the consistency of operations across multiple robotic arms in a unified space, it is usually necessary to introduce spatial calibration methods based on vision or pose information, as well as hand-eye calibration mechanisms, to establish spatial mapping relationships between robotic arms and workpieces, and among multiple robotic arms, thereby supporting the basic control for collaborative operations.

[0003] Traditional heterogeneous dual-arm calibration and error compensation mostly adopt static calibration models and fixed coordinate transformation relationships. Since the workpiece spatial state and the pose changes of multiple robotic arms are not uniformly constrained and modeled, problems such as mismatch of cross-device spatial mapping relationship and error accumulation are caused. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms. It aims to improve the problem that traditional heterogeneous dual robotic arm calibration and error compensation mostly adopt static calibration models and fixed coordinate transformation relationships, which easily cause mismatch of cross-device spatial mapping relationships and error accumulation.

[0005] This invention provides the following technical solution: a method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms, comprising the following steps: S1. Obtain the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data. Perform time synchronization and spatial alignment on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data, and generate collaborative spatial state data. S2. Based on the collaborative spatial state data and the dynamic constraint state applied to the workpiece by the first heterogeneous robotic arm, a workpiece-level dynamic process reference constraint system is constructed, and the workpiece-level dynamic process reference constraint system is updated according to the workpiece holding state, workpiece deformation and workpiece spatial drift. S3. Based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece-level dynamic process reference constraint system, establish a reference transfer chain and reconstruct the collaborative space mapping relationship. S4. Based on the collaborative space mapping relationship, update the dynamic hand-eye mapping relationship between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system. S5. Generate the machining trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship. When a change in the workpiece spatial state or a change in the relative pose of the first heterogeneous robotic arm and the second heterogeneous robotic arm is detected, update the workpiece-level dynamic process reference constraint system, reference transfer chain, collaborative space mapping relationship and dynamic hand-eye mapping relationship, and perform correction on the machining trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship.

[0006] By adopting the above technical solution, the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data, are integrated under the same collaborative spatial state to establish a dynamically updated workpiece-level dynamic process reference constraint system. This enables the reconstruction of cross-robotic arm collaborative spatial mapping relationship and dynamic hand-eye mapping adaptive update based on the reference transfer chain. This improves the problem that traditional heterogeneous dual-robotic arm calibration and error compensation mostly use static calibration models and fixed coordinate transformation relationships. Since the workpiece spatial state and the pose changes of multiple robotic arms are not uniformly constrained and modeled, the mismatch of cross-equipment spatial mapping relationship and error accumulation are caused.

[0007] The present invention has the following beneficial effects: 1. In this invention, by fusing real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as workpiece spatial state and visual observation data, and establishing a dynamically updated workpiece-level dynamic process reference constraint system under the same collaborative spatial state, the reconstruction of cross-robotic arm collaborative spatial mapping relationship and dynamic hand-eye mapping adaptive update based on the reference transfer chain are realized. This improves the problem that traditional heterogeneous dual-robotic arm calibration and error compensation mostly adopt static calibration models and fixed coordinate transformation relationships. Since the workpiece spatial state and the pose changes of multiple robotic arms are not uniformly constrained and modeled, the mismatch of cross-equipment spatial mapping relationship and error accumulation are caused.

[0008] 2. In this invention, the dynamic process reference constraint system at the workpiece level is continuously updated based on the workpiece holding state, workpiece deformation, and workpiece spatial drift, and the reference direction field and dynamic process reference structure of the workpiece processing area are simultaneously corrected. This enables the dynamic reconstruction of the processing reference as the workpiece state changes, thereby improving the traditional method of establishing processing reference, which mostly relies on the initial workpiece geometric model and fixed processing coordinate reference. Due to the lack of continuous response to the deformation and spatial drift of the workpiece during the holding process, the processing reference offset accumulates.

[0009] 3. In this invention, a second heterogeneous robotic arm machining trajectory is generated based on a dynamic hand-eye mapping relationship. When a change in the workpiece's spatial state or a change in the relative pose of the robotic arm is detected, the workpiece-level dynamic process reference constraint system, reference transfer chain, and collaborative space mapping relationship are updated in conjunction with the machining trajectory, and real-time correction is performed on the machining trajectory. This achieves synchronous adaptive adjustment of the machining trajectory and changes in spatial state, thereby improving the problem that traditional robotic arm machining trajectory control mostly adopts pre-planned trajectory and offline error compensation methods. Due to the difficulty in responding to changes in the workpiece state and the relative pose of multiple robotic arms in a timely manner, trajectory deviation and machining consistency are reduced. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the large-scale spatial dynamic calibration and hand-eye error compensation method for heterogeneous dual robotic arms proposed in this invention. Figure 2 This is a schematic diagram of the architecture of the heterogeneous dual robotic arm large-scale spatial dynamic calibration and hand-eye error compensation system proposed in an embodiment of the present invention. Detailed Implementation

[0011] The technical solutions in 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.

[0012] Example 1: In the first embodiment of the present invention, a method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms is provided, such as... Figure 1 As shown, the process includes the following steps: S1, acquiring real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as workpiece spatial state and visual observation data, performing time synchronization and spatial alignment on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as workpiece spatial state and visual observation data, to generate collaborative spatial state data. Furthermore, in S1, the steps of synchronizing and aligning the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data, with time include: Timestamp matching is performed on the real-time pose data of the first heterogeneous robotic arm and the real-time pose data of the second heterogeneous robotic arm based on a unified time reference sequence. The observation time is aligned between the workpiece spatial state and the visual observation data based on the visual acquisition timestamp. Establish the spatial transformation relationship between the first heterogeneous robot arm base coordinate system, the second heterogeneous robot arm base coordinate system, and the workpiece local coordinate system; Based on spatial transformation relationships, coordinate unification is performed on the real-time pose data of the first heterogeneous robotic arm, the real-time pose data of the second heterogeneous robotic arm, and the workpiece spatial state and visual observation data to generate collaborative spatial state data.

[0013] Specifically, by acquiring real-time pose data of the first and second heterogeneous robotic arms, as well as workpiece spatial state and visual observation data, and using a unified time reference sequence as a time base to achieve time synchronization processing of multi-source data, and simultaneously completing cross-coordinate system integration based on spatial transformation relationships, collaborative spatial state data for subsequent collaborative modeling is generated. The real-time pose data of the first heterogeneous robotic arm is denoted as... The real-time pose data of the second heterogeneous robotic arm is recorded as follows: The spatial state of the workpiece is denoted as , where X W (t)={P W (t),T W (t)} is used to represent a unified expression of the geometric state of the workpiece point cloud and the global pose state of the workpiece. The visual observation data is denoted as The unified time reference series is denoted as The time synchronization process can be represented as mapping data from different sources to the same time index relationship. Thus, the synchronized data set is obtained. Timestamp matching is achieved by defining a time deviation function. Achieve minimum time offset alignment and select the option that satisfies... The sampled frames are used as the synchronization result, where t A The timestamp for the real-time pose data of the first heterogeneous robotic arm is t. B The timestamp for the real-time pose data of the second heterogeneous robotic arm is t. V The hardware trigger timestamps for image acquisition by the vision system are all obtained by synchronizing their respective controller clocks using a network time protocol. In this scheme, the time synchronization accuracy is preferably controlled within the millisecond sampling interval range. After time alignment is completed, a first heterogeneous robotic arm base coordinate system ∑ is established. A The second heterogeneous robotic arm base coordinate system ∑ TCP and the workpiece local coordinate system ∑ W Spatial alignment is achieved through spatial transformation relationships, where the spatial transformation relationships are represented by homogeneous transformation matrices. and The visual observation coordinate system is denoted as ∑ V And obtained through external parameter calibration This allows for the construction of a unified coordinate representation system. Based on this system, various state variables undergo unified coordinate transformation processing. For example, mapping the end-effector pose to the workpiece coordinate system can be represented as follows: Similarly, visual observation data is mapped to ;in , , These represent the spatial positions of points in their respective coordinate systems. In this scheme, the homogeneous transformation matrix is ​​preferably based on a rotation matrix. With translation vector Form of composition The rotation matrix describes the attitude relationship, the translation vector describes the spatial position relationship, and finally all the unified spatial representations are aggregated into collaborative spatial state data. = (T A (t i ),T B (t i ),X W (t i ),X V (t i This result is used to subsequently construct a workpiece-level dynamic process reference constraint system and a collaborative space mapping relationship, thereby providing a unified spatial input basis for multi-robotic arm collaborative control and hand-eye error compensation.

[0014] S2. Based on the collaborative spatial state data and the dynamic constraint state applied to the workpiece by the first heterogeneous robotic arm, a workpiece-level dynamic process reference constraint system is constructed, and the workpiece-level dynamic process reference constraint system is updated according to the workpiece holding state, workpiece deformation and workpiece spatial drift. Furthermore, in S2, the steps for constructing a workpiece-level dynamic process reference constraint system include: Based on workpiece spatial state and visual observation data, extract workpiece surface features, curved surface edge contours, and curved surface normal distribution; The key machining areas of the workpiece are determined based on the workpiece surface features and spatial geometric constraints. Dynamic process reference point set, dynamic process reference edge and dynamic process reference surface are generated based on the key processing area of ​​the workpiece. A reference direction field for the workpiece machining area is constructed based on the surface normal distribution and dynamic process reference surface. A workpiece-level dynamic process reference constraint system is generated based on a dynamic process reference point set, dynamic process reference edges, dynamic process reference surfaces, and a workpiece machining area reference direction field.

[0015] Specifically, by using collaborative spatial state data and the dynamic constraint state applied to the workpiece by the first heterogeneous robotic arm as joint inputs, a workpiece-level dynamic process reference constraint system is constructed. Based on visual observation data, multi-scale feature extraction of the workpiece geometry is performed, and a unified constraint expression is established, where the workpiece spatial state is denoted as X. W (t)={P W(t),T W (t)}, where P W (t) represents the geometric state of the workpiece point cloud, T W (t) represents the global pose state of the workpiece; the visual observation data is denoted as X. V (t); First, by fusing X W The point cloud in (t) and X V The observed point cloud in (t) is reconstructed into a surface to obtain the workpiece point cloud set P′ used for modeling. W ={pi}; where Indicates the first Each spatial sampling point is obtained from the output data of a vision sensor after temporal and spatial synchronization; based on this, the surface normal distribution is calculated through spatial gradient estimation. Its calculation method is as follows ;in This represents a surface function obtained based on local neighborhood fitting. This represents the spatial gradient vector at that point. In this scheme, the neighborhood range is preferably represented by the nearest neighbor. Fitting is performed on 10 sampling points, where A range of 20 to 50 can be selected to balance computational stability and local accuracy; subsequently, boundary contours are extracted based on the workpiece point cloud set and normal distribution information to form a set of surface edge contours. and through space constraint functions The set of critical processing areas is obtained by screening areas that meet the processing conditions. The constraint function can be expressed as ;in This represents the distance feature from a point to the geometric center of the workpiece. Indicates the angle between the normal and the reference direction. and For the weighting coefficients, the preferred solution in this scheme is one that satisfies... and Values ​​between 0.4 and 0.6 are used to balance spatial distribution and directional consistency; after determining the key processing areas, a dynamic process reference point set is generated through regional sampling. Dynamic process reference edge and dynamic process reference surface The reference point set is obtained by means of... The reference edges are obtained by uniform sampling based on spatial density, and the reference surfaces are obtained by curve fitting of the boundary points. The general form is as follows: ; parameters Derived from the normal mean direction estimation results; further based on the surface normal distribution With reference surface set Constructing a reference orientation field for the processing area Its definition is: D W (p i )=n i Among them, for those falling on the dynamic process reference plane S F For points within the neighborhood, set their normal vector n. i Corrected to reference plane normal n ref This involves applying consistency constraints to the orientation field of a local region to reflect the standardizing role of the process reference in the orientation of the surface; ultimately, the dynamic process reference point set, dynamic process reference edge, dynamic process reference surface, and reference orientation field are uniformly integrated to construct a workpiece-level dynamic process reference constraint system. Essentially, it is a set of constraints composed of spatial geometric constraints, directional constraints, and structural constraints. It is used as a unified spatial constraint input for updating the reference transfer chain and hand-eye mapping relationship in subsequent steps, thereby supporting the dynamic consistency maintenance and continuous updating of the spatial reference in the collaborative processing of the robotic arm.

[0016] Furthermore, in S2, the steps for updating the workpiece-level dynamic process reference constraint system based on the workpiece holding state, workpiece deformation, and workpiece spatial drift include: The changes in workpiece holding posture, workpiece deformation, workpiece spatial drift, and surface normal are calculated based on the workpiece spatial state and visual observation data at continuous time intervals. The offset of reference elements in the workpiece-level dynamic process datum constraint system is calculated based on the workpiece holding posture change, workpiece deformation, and workpiece spatial drift. Update the dynamic process reference point set, dynamic process reference edge, and dynamic process reference surface based on the reference feature offset; Update the reference orientation field of the workpiece machining area based on the change in surface normal; The workpiece-level dynamic process reference constraint system is updated based on the updated dynamic process reference point set, dynamic process reference edge, dynamic process reference surface, and reference direction field.

[0017] Specifically, time-varying correction of the workpiece-level dynamic process reference constraint system is achieved through continuous workpiece spatial state and visual observation data. The core implementation method involves uniformly modeling the changes in the workpiece's spatial geometric state and attitude based on multi-source time-series observations, and mapping these changes to the structural parameter updates of the reference constraint system. The workpiece spatial state is denoted as... Visual observation data is denoted as First, a state difference model is constructed based on data from adjacent time points. ;in This indicates the spatial observation state of the fused workpiece. The sampling time interval is represented by a fixed period sampling method preferred in this scheme to ensure time continuity. ΔX(t) is mapped to the workpiece holding posture change, workpiece deformation, workpiece spatial drift, and surface normal change through a state analysis function, expressed as: (ΔP h ,ΔP d ,ΔP s ,ΔN)=Φ(ΔX(t)); where Φ represents the state decomposition mapping function; based on this difference model, the changes in the workpiece holding posture are extracted respectively. Workpiece deformation Workpiece spatial drift and the change in surface normal The change in the holder's posture is obtained through rigid body pose decomposition and can be expressed as: in The current moment's workpiece attitude homogeneous transformation matrix is ​​represented by the deformation variables obtained through the Euclidean distance changes between corresponding points in the point cloud. The spatial drift is calculated by the change in the overall centroid of the workpiece, and is expressed as... ;in The change in the surface normal is calculated by the change in the angle between the normal vectors, where the centroid of the workpiece point cloud is located. This is expressed as ΔN = arccos(n i (t) n i (t-Δt)); where The current normal direction is n, where n is the current normal direction. i Let be the normal vector of the i-th sampling point; after obtaining the above changes, a mapping function is constructed. The change is mapped to the offset of the reference element in the workpiece-level dynamic process datum constraint system. ,Right now In this scheme, the mapping function preferably adopts a weighted combination form. The weighting coefficients satisfy the following conditions: This is used to represent the degree of impact of different types of changes on the baseline structure; subsequently, the dynamic process reference point set is adjusted based on the reference element offset. Dynamic process reference edge and dynamic process reference surface Updates are performed, with point set updates achieved through... The update of the dynamic process reference edge is achieved by translating its associated control points by ΔB and then refitting the curve equation; the update of the dynamic process reference surface is achieved by translating the point set of the surface definition by ΔB and then recalculating the plane or surface parameters, while simultaneously adjusting the reference direction field of the processing area based on the change in the surface normal. To update, it is expressed as The orientation field is updated point-by-point with spatial point p as the independent variable to reflect local orientation offset, where ΔN(p) is calculated through local neighborhood normal difference. Finally, the updated reference point set, reference edge, reference surface, and orientation field are re-integrated to construct a new workpiece-level dynamic process reference constraint system. This result is used to provide a basis for real-time updated spatial constraints in subsequent reference transfer chains and collaborative space mapping relationships, thereby supporting continuous response to changes in workpiece state and adaptive error correction during multi-robot collaborative processing.

[0018] S3. Based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece-level dynamic process reference constraint system, establish a reference transfer chain and reconstruct the collaborative space mapping relationship. Furthermore, in S3, the steps for establishing the reference transfer chain include: Based on the real-time pose data of the first heterogeneous robotic arm, establish the spatial transformation relationship between the base coordinate system of the first heterogeneous robotic arm and the end-effector coordinate system of the first heterogeneous robotic arm; Based on the workpiece-level dynamic process reference constraint system and workpiece spatial state and visual observation data, establish the spatial transformation relationship between the first heterogeneous robotic arm end coordinate system and the workpiece-level dynamic process reference constraint system. Based on the real-time pose data of the second heterogeneous robotic arm and the spatial state and visual observation data of the workpiece, a spatial transformation relationship is established between the workpiece-level dynamic process reference constraint system and the machining coordinate system of the second heterogeneous robotic arm. Based on the above spatial transformation relationship, a reference transfer chain is established from the workpiece-level dynamic process reference constraint system to the machining coordinate system of the second heterogeneous robotic arm.

[0019] Specifically, a reference transfer chain is formed by constructing a continuous spatial transformation relationship between multiple coordinate systems. This is achieved by unifying the coordinate representation of the real-time pose data of the first and second heterogeneous robotic arms with the workpiece-level dynamic process reference constraint system, and establishing an intermediate spatial bridging relationship through visual observation data. This forms a complete transfer path from the robotic arm base coordinate system to the machining coordinate system, where the first heterogeneous robotic arm base coordinate system is represented as ∑ A The coordinate system of the end effector of the first heterogeneous robotic arm is represented as ∑ AE The second heterogeneous robotic arm tool coordinate system is represented as ∑ TCP The workpiece-level dynamic process datum constraint system corresponds to the spatial coordinate system represented as ∑ W The visual observation coordinate system is represented as ∑ V The real-time pose data of the first heterogeneous robotic arm is represented by a homogeneous transformation matrix. The real-time pose data of the second heterogeneous robotic arm is represented as follows: The spatial state of the workpiece and the visual observation data are mapped to a unified observation state after time synchronization. Define the second heterogeneous robotic arm machining coordinate system ∑ TCP Let the coordinate system be the tool center point fixed at the end of the second heterogeneous robotic arm, and its coordinate system relative to the end of the second heterogeneous robotic arm be ∑. BE Transformation relationship Obtained from tool parameter calibration, and considered a known constant in this scheme, the spatial transformation relationship from the base coordinate system to the end effector coordinate system is first constructed from the pose data of the first heterogeneous robotic arm, and expressed as follows: This transformation is calculated using the forward kinematics model of the robotic arm, where the forward kinematics input is the joint angle vector. The output is the end pose matrix. In this scheme, the joint angle data is acquired in real time by the robot arm encoder; subsequently, the transformation relationship from the end-effector coordinate system to the workpiece reference coordinate system is established through the workpiece spatial state and visual observation data, expressed as follows: The rigid body transformation matrix f is obtained by least-squares point cloud registration, and its input is the workpiece point cloud set. With end pose observation point set Among them, the set of end-effector pose observation points The acquisition method is as follows: A visual marker pre-calibrated at the end of the first heterogeneous robotic arm is used, and its pose is observed by an external vision system and transformed to the workpiece coordinate system. Alternatively, a vision sensor fixed to the end of the first heterogeneous robotic arm directly observes the workpiece reference marker and acquires the coordinates through hand-eye relationship transformation. The output is a rigid body transformation matrix. ;in Represents the rotation matrix. The translation vector is represented by the least squares point cloud registration method, which is preferred to be solved in this scheme. Similarly, the spatial transformation relationship between the workpiece coordinate system and the machining coordinate system of the second robotic arm is established by using the real-time pose data of the second heterogeneous robotic arm and the visual observation data. The spatial transformation relationship is solved based on the iterative nearest-point registration method. Its input includes the workpiece point cloud and the pose observation of the second robotic arm's end effector, and the output is a homogeneous transformation matrix. ;in and Let these represent spatial rotation and translation relationships, respectively. In this scheme, the iterative nearest-point registration algorithm is preferably used to implement this process. After obtaining the above three transformation relationships, a complete reference transfer chain is established through matrix chain multiplication, as shown below. This expression realizes a unified spatial mapping output from the base coordinate system of the first robotic arm to the machining coordinate system of the second robotic arm. The result is used for subsequent collaborative spatial mapping relationship reconstruction, enabling different robotic arms to achieve spatial consistency expression and error transmission control under the same workpiece dynamic reference, thereby providing a unified geometric basis for subsequent dynamic hand-eye mapping updates.

[0020] Furthermore, in S3, the steps for reconstructing the cooperative space mapping relationship include: The relative spatial transformation relationship between the first heterogeneous robotic arm and the second heterogeneous robotic arm is calculated based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm. Based on the workpiece-level dynamic process reference constraint system and workpiece spatial state and visual observation data, the workpiece spatial constraint relationship is extracted. Establish a mapping relationship between the workpiece-level dynamic process datum constraint system and the visual coordinate system based on the workpiece-level dynamic process datum constraint system; Based on the workpiece-level dynamic process reference constraint system and the relative spatial transformation relationship between the first heterogeneous robotic arm and the second heterogeneous robotic arm, a mapping relationship is established from the workpiece-level dynamic process reference constraint system to the machining coordinate system of the second heterogeneous robotic arm. Based on the above mapping relationship and the relative spatial transformation relationship between the first heterogeneous robotic arm and the second heterogeneous robotic arm, a collaborative spatial mapping relationship is generated.

[0021] Specifically, by jointly modeling the relative pose relationship between the two robotic arms and the workpiece-level dynamic process reference constraint system, a unified spatial mapping relationship across equipment is reconstructed. This is achieved by performing a relative transformation on the real-time pose data of the first and second heterogeneous robotic arms, and combining this with visual observation data to apply geometric consistency constraints to the workpiece spatial constraints. This constructs a bidirectional mapping system from the workpiece reference to both the visual coordinate system and the machining coordinate system. The real-time pose of the first heterogeneous robotic arm is represented as follows: The real-time pose of the second heterogeneous robotic arm is represented as follows: The workpiece-level dynamic process datum constraint system is represented as: Visual observation data is represented as First, the spatial relationship between the first heterogeneous robotic arm and the second heterogeneous robotic arm is calculated through relative transformation and expressed as follows: This transformation originates from the matrix difference calculation of the end-effector poses of the two robotic arms under a unified time reference sequence. The pose matrix is ​​calculated from the joint angle vectors using a forward kinematics model. In this scheme, the joint angles are derived from real-time sampling data from the robotic arm encoder. Subsequently, based on the workpiece-level dynamic process reference constraint system and visual observation data, the workpiece spatial constraint relationship is extracted and expressed as follows: ;where the function This represents the spatial constraint mapping based on the solution of point cloud geometric consistency and surface constraint consistency. It calculates spatial constraint relationships based on point cloud geometric consistency constraints, and its input includes the workpiece point cloud set. With visual observation point set The output is a set of spatial constraints. This constraint set is used to describe the local geometric stability of the workpiece and its consistency with the reference structure; based on this, the mapping relationship between the workpiece-level dynamic process reference constraint system and the visual coordinate system is established as follows: ;in and The coordinates are obtained from the visual point cloud registration results. In this scheme, the registration method based on nearest neighbor iteration optimization is preferred. Furthermore, the mapping relationship between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system is established by combining the workpiece spatial constraint relationship and the relative pose relationship. This expression is used to fuse the visual reference with the relative motion relationship of the robotic arm, thereby achieving a cross-coordinate system mapping; finally, a cooperative spatial mapping relationship is generated by fusing the mapping relationship with the relative pose relationship, represented as follows: ;where the function The spatial consistency fusion operator outputs a unified collaborative mapping matrix, which is used for subsequent dynamic hand-eye mapping relationship updates and machining trajectory corrections, thereby enabling the dual robotic arms to achieve a consistent spatial expression and collaborative control foundation under the workpiece dynamic reference constraint system.

[0022] S4. Based on the collaborative space mapping relationship, update the dynamic hand-eye mapping relationship between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system. Furthermore, in S4, the step of updating the dynamic hand-eye mapping relationship between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system includes: The offset of reference elements in the workpiece-level dynamic process datum constraint system is calculated based on the workpiece-level dynamic process datum constraint system and the collaborative space mapping relationship. The relative pose change is calculated based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm. The surface space deviation is calculated based on the collaborative spatial mapping relationship and the workpiece spatial state and visual observation data. The dynamic hand-eye mapping relationship is updated based on the offset, relative pose change, and surface space deviation of the reference elements in the workpiece-level dynamic process datum constraint system.

[0023] Specifically, by constructing a collaborative spatial mapping relationship as a unified coordinate reference, the spatial correspondence between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system is transformed into a time-varying hand-eye mapping relationship update problem. This is achieved by uniformly modeling and solving three error sources: workpiece-side reference drift, relative motion of the robotic arm, and visual observation deviation, thereby realizing dynamic correction of the mapping relationship. The specific process is as follows: First, the reference element offset is calculated based on the workpiece-level dynamic process reference constraint system and the collaborative spatial mapping relationship. The spatial representation of the workpiece-level dynamic process reference constraint system at time t is assumed to be in matrix form. The collaborative spatial mapping relationship is represented by the transformation matrix from the visual coordinate system to the workpiece reference coordinate system. Then the theoretical position of the reference feature in the workpiece datum is ;in This represents the three-dimensional coordinate vector of workpiece feature points obtained from visual observation. In this scheme, these coordinates are extracted by binocular vision or structured light point cloud, preferably centimeter-scale normalized coordinates. The reference element offset is defined as... ;in The coordinates of the reference point for workpiece design are used; secondly, the relative pose change is calculated based on the real-time pose data of the first and second heterogeneous robotic arms, assuming the end-effector pose of the first robotic arm is a homogeneous transformation matrix. The end effector pose of the second robotic arm is The relative pose relationship is defined as follows: The pose data is derived from the fusion calculation of the robotic arm encoder and the forward kinematics model. In this scheme, the position component accuracy is preferably at the millimeter level, and the attitude components are expressed in Euler angles or quaternions. Furthermore, the surface space deviation is calculated based on the cooperative spatial mapping relationship, the workpiece spatial state, and visual observation data. Let the workpiece surface observation point cloud be... Its corresponding fitted surface is Then the surface normal deviation can be expressed as Meanwhile, the surface position deviation is expressed as ;in Represents the normal estimation function, Let represent the point-to-surface distance function, both of which are obtained by local surface fitting and normal estimation algorithms; finally, the reference feature offset, relative pose change, and surface space deviation are fused to update the dynamic hand-eye mapping relationship, assuming the dynamic hand-eye mapping relationship is denoted as . The update process can then be represented as In the above formula, log( ) is a matrix logarithmic mapping used to map a homogeneous transformation matrix to a Lie algebra vector; exp( ) is a matrix exponential mapping used to map Lie algebra vectors back to homogeneous transformation matrix increments. The specific update process is as follows: first, the relative pose change log(T) is... rel,t Translation components and reference element offset Δp in ) ref,t A weighted fusion is performed to generate a total pose deviation Lie algebra vector, which is then combined with the surface space deviation Δd. t The constructed orientation-corrected Lie algebra vectors are merged and jointly generated through exponential mapping to produce an update increment matrix, thereby achieving dynamic correction of the hand-eye mapping relationship.

[0024] S5. Generate the machining trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship. When a change in the workpiece spatial state or a change in the relative pose of the first heterogeneous robotic arm and the second heterogeneous robotic arm is detected, update the workpiece-level dynamic process reference constraint system, reference transfer chain, collaborative space mapping relationship and dynamic hand-eye mapping relationship, and perform correction on the machining trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship. Furthermore, in S5, the step of correcting the machining trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship includes: The trajectory offset of the workpiece machining area is calculated based on the dynamic hand-eye mapping relationship and the workpiece-level dynamic process reference constraint system. The workpiece machining area posture offset is calculated based on the workpiece spatial state, visual observation data, and dynamic hand-eye mapping relationship. The corrected machining trajectory of the second heterogeneous robotic arm is generated based on the trajectory offset and attitude offset. The current processing path of the second heterogeneous robotic arm is updated based on the corrected processing trajectory.

[0025] Specifically, by applying the spatial transformation capability output by the dynamic hand-eye mapping relationship to the geometric reference structure in the workpiece-level dynamic process datum constraint system, and combining it with real-time feedback from visual observation, the adaptive correction of the machining trajectory of the second heterogeneous robotic arm under spatial offset and attitude drift conditions is achieved. This is accomplished by calculating the trajectory position error and attitude error separately within a unified mapping coordinate framework, then synthesizing and reconstructing the trajectory and writing back the control path, thus forming a closed-loop compensation process. First, the trajectory offset of the workpiece machining area is calculated based on the dynamic hand-eye mapping relationship and the workpiece-level dynamic process datum constraint system. Let the dynamic hand-eye mapping relationship be represented by a homogeneous transformation matrix. In the workpiece-level dynamic process datum constraint system, the reference trajectory points of the machining area are represented as: Then the theoretical trajectory point of the second heterogeneous robotic arm at the current moment is: ;in Derived from the dynamic process reference point set and the discrete sampling results of the reference edge, the sampling interval is preferably at the millimeter level in this scheme to ensure trajectory continuity. The trajectory offset is defined as... ;in The original planned trajectory points for the second heterogeneous robotic arm are generated by the path planning module based on the processing task. Next, the workpiece's attitude offset in the processing area is calculated based on the workpiece's spatial state, visual observation data, and the dynamic hand-eye mapping relationship. Let the local workpiece attitude obtained from visual observation be the rotation matrix. The pose components in the dynamic hand-eye mapping relationship are: The theoretical processing posture is Attitude offset is defined as ;in The original machining trajectory corresponds to the posture, which is obtained by inverse kinematics calculation by the robotic arm controller. Further, a corrected second heterogeneous robotic arm machining trajectory is generated based on the trajectory offset and posture offset, unifying the position and posture into a homogeneous matrix form. The corrected trajectory is defined as: In this scheme, the process is made continuous through trajectory interpolation and smoothing filtering to avoid abrupt changes; finally, the current processing path of the second heterogeneous robotic arm is updated based on the corrected processing trajectory. The input is sent to the robotic arm motion control module as the control command for the next cycle, so that the processing path is adjusted in real time according to the changes in the workpiece spatial state and the relative motion of the two robotic arms, realizing continuous consistency correction of the trajectory under the conditions of spatial offset and posture error.

[0026] It also includes the following steps: Calculate the system state consistency index based on workpiece spatial state and visual observation data; Dynamic process baseline reliability parameters are generated based on system state consistency indicators. When the reliability parameter of the dynamic process reference is lower than the preset threshold, the reconstruction of the workpiece-level dynamic process reference constraint system, reference transfer chain and collaborative space mapping relationship is triggered.

[0027] Specifically, by constructing a system state consistency evaluation mechanism, the consistency between visual observation results and workpiece spatial state is measured under a unified spatial benchmark. This consistency result is used as a dynamic process benchmark reliability criterion, thereby driving the adaptive reconstruction of the process benchmark system and spatial mapping relationship. This enables the calibration and compensation system to achieve self-recovery under abnormal drift conditions. The implementation method is to calculate the consistency index based on multi-source state error fusion, form a reliability parameter through normalized mapping, and then trigger the system reconstruction process through threshold judgment. First, the system state consistency index is calculated based on the workpiece spatial state and visual observation data. Let the workpiece spatial state obtained by visual observation be a point cloud set. The theoretical spatial state derived from the workpiece-level dynamic process datum constraint system is a set. This set is obtained by mapping the set of feature points on the workpiece design model to the current workpiece coordinate system after non-rigid body registration under the constraints of dynamic process reference point sets, reference surfaces, and other elements in the reference constraint system. The consistency error is then defined as... ;in Indicates the visual observation of the first Coordinates of feature points This indicates the coordinates of the theoretical points corresponding to the reference system. The number of feature points in this scheme Ideally, there should be no fewer than twenty to ensure statistical stability. The system state consistency index is defined as follows: ;in The error attenuation coefficient is used to adjust the intensity of the error's impact on the consistency index, and in this scheme, it is preferably a positive real constant. Secondly, a dynamic process baseline reliability parameter is generated based on the system state consistency index, and the reliability parameter is defined as a time-weighted consistency function. ;in This represents the confidence parameter at the current moment. This represents the confidence parameter at the previous time step. To update the weighting coefficients and balance the contributions of historical states and current observations, this scheme preferably uses continuous values ​​between zero and one. Finally, when the dynamic process baseline confidence parameter falls below a preset threshold, system reconstruction is triggered, with the following judgment condition: , As a preset threshold parameter, the preferred stability threshold in this scheme is set according to the initial error level of the calibration. Once the condition is triggered, the workpiece-level dynamic process reference constraint system reconstruction, reference transfer chain reconstruction, and collaborative spatial mapping relationship reconstruction are performed. The reconstruction process includes re-executing visual feature extraction, coordinate system relationship re-estimation, and spatial transformation chain update, thereby restoring the overall consistency and mapping stability of the system under workpiece spatial drift or sudden changes in the relative pose of the two robotic arms, and ensuring that subsequent machining trajectory correction is calculated based on the updated reliable spatial reference.

[0028] Example 2: In the second embodiment of the present invention, the present invention provides a large-scale spatial dynamic calibration and hand-eye error compensation system for heterogeneous dual robotic arms, such as... Figure 2 As shown, it includes the following modules: The data synchronization module is used to acquire the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data. It performs time synchronization and spatial alignment on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data, and generates collaborative spatial state data. The benchmark construction module is used to construct a workpiece-level dynamic process benchmark constraint system based on collaborative space state data and the dynamic constraint state applied to the workpiece by the first heterogeneous robotic arm, and to update the workpiece-level dynamic process benchmark constraint system according to the workpiece holding state, workpiece deformation and workpiece spatial drift. The link mapping module is used to establish a reference transfer chain and reconstruct the collaborative space mapping relationship based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece-level dynamic process reference constraint system. The mapping update module is used to update the dynamic hand-eye mapping relationship between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system based on the collaborative space mapping relationship. The trajectory correction module is used to generate the machining trajectory of the second heterogeneous robot arm based on the dynamic hand-eye mapping relationship. When a change in the workpiece spatial state or a change in the relative pose of the first heterogeneous robot arm and the second heterogeneous robot arm is detected, the workpiece-level dynamic process reference constraint system, reference transfer chain, cooperative space mapping relationship and dynamic hand-eye mapping relationship are updated, and the machining trajectory of the second heterogeneous robot arm is corrected based on the dynamic hand-eye mapping relationship.

[0029] In large-scale heterogeneous dual-manipulator collaborative machining scenarios, such as the fixtureless assembly of an aero-engine casing and a large-size thin-walled shell, traditional static calibration methods struggle to maintain spatial consistency due to the separation of the dual-manipulator mounting base, insufficient workpiece rigidity, deformation during machining, and visual measurement drift. This easily leads to trajectory deviation and accumulation of collaborative errors. To address these issues, the large-scale spatial dynamic calibration and hand-eye error compensation system for heterogeneous dual-manipulators provided in this invention is employed. Its structure is as follows: Figure 2 As shown. The specific implementation process of this system is as follows: First, the real-time pose data of the two robotic arms, as well as the spatial state and visual observation data of the workpiece, are acquired through the data synchronization module. Time synchronization and spatial alignment are then performed to eliminate the time deviation and coordinate inconsistency of multi-source data, generating unified collaborative spatial state data to provide consistent input for subsequent calculations. Then, the reference construction module constructs a workpiece-level dynamic process reference constraint system based on the collaborative space state data and the dynamic constraint state of the first robotic arm on the workpiece. It continuously updates the reference system by combining the workpiece holding state, deformation and spatial drift, so that the workpiece reference can be adaptively adjusted with the processing process, thereby improving the reference stability under the condition of flexible workpiece. Next, a reference transfer chain is established based on the real-time pose data of the two robotic arms and the dynamic process reference constraint system through the link mapping module, and the collaborative spatial mapping relationship is reconstructed to realize the unified spatial expression between different coordinate systems, thereby ensuring the spatial consistency of the two robotic arms under cross-base conditions. Subsequently, the dynamic hand-eye mapping relationship between the workpiece datum and the second robotic arm machining coordinate system is updated based on the collaborative space mapping relationship through the mapping update module. Workpiece drift, robotic arm motion and visual errors are uniformly incorporated into the update model to achieve online correction and continuous compensation of the hand-eye relationship. Finally, the trajectory correction module generates a machining trajectory based on the dynamic hand-eye mapping relationship. When a change in the workpiece state or a change in the relative pose of the two robotic arms is detected, the reference system and mapping relationship are updated in conjunction with the trajectory correction module, and the machining trajectory is corrected in real time. This ensures that the second robotic arm always performs machining under the adaptively updated spatial reference, thereby achieving closed-loop compensation for collaborative errors and stable trajectory control.

[0030] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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 large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms, characterized in that, Includes the following steps: S1. Obtain the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data. Perform time synchronization and spatial alignment on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data, and generate collaborative spatial state data. S2. Based on the collaborative spatial state data and the dynamic constraint state applied to the workpiece by the first heterogeneous robotic arm, a workpiece-level dynamic process reference constraint system is constructed, and the workpiece-level dynamic process reference constraint system is updated according to the workpiece holding state, workpiece deformation and workpiece spatial drift. S3. Based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece-level dynamic process reference constraint system, establish a reference transfer chain and reconstruct the collaborative space mapping relationship. S4. Based on the collaborative space mapping relationship, update the dynamic hand-eye mapping relationship between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system. S5. Generate the machining trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship. When a change in the workpiece spatial state or a change in the relative pose of the first heterogeneous robotic arm and the second heterogeneous robotic arm is detected, update the workpiece-level dynamic process reference constraint system, reference transfer chain, collaborative space mapping relationship and dynamic hand-eye mapping relationship, and perform correction on the machining trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship.

2. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, In S1, the step of synchronizing and spatially aligning the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm, as well as the workpiece spatial state and visual observation data, includes: Timestamp matching is performed on the real-time pose data of the first heterogeneous robotic arm and the real-time pose data of the second heterogeneous robotic arm based on a unified time reference sequence. The observation time is aligned between the workpiece spatial state and the visual observation data based on the visual acquisition timestamp. Establish the spatial transformation relationship between the first heterogeneous robot arm base coordinate system, the second heterogeneous robot arm base coordinate system, and the workpiece local coordinate system; Based on spatial transformation relationships, coordinate unification is performed on the real-time pose data of the first heterogeneous robotic arm, the real-time pose data of the second heterogeneous robotic arm, and the workpiece spatial state and visual observation data to generate collaborative spatial state data.

3. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, In S2, the step of constructing the workpiece-level dynamic process reference constraint system includes: Based on workpiece spatial state and visual observation data, extract workpiece surface features, curved surface edge contours, and curved surface normal distribution; The key machining areas of the workpiece are determined based on the workpiece surface features and spatial geometric constraints. Dynamic process reference point set, dynamic process reference edge and dynamic process reference surface are generated based on the key processing area of ​​the workpiece. A reference direction field for the workpiece machining area is constructed based on the surface normal distribution and dynamic process reference surface. A workpiece-level dynamic process reference constraint system is generated based on a dynamic process reference point set, dynamic process reference edges, dynamic process reference surfaces, and a workpiece machining area reference direction field.

4. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, In S2, the step of updating the workpiece-level dynamic process reference constraint system based on the workpiece holding state, workpiece deformation, and workpiece spatial drift includes: The changes in workpiece holding posture, workpiece deformation, workpiece spatial drift, and surface normal are calculated based on the workpiece spatial state and visual observation data at continuous time intervals. The offset of reference elements in the workpiece-level dynamic process datum constraint system is calculated based on the workpiece holding posture change, workpiece deformation, and workpiece spatial drift. Update the dynamic process reference point set, dynamic process reference edge, and dynamic process reference surface based on the reference feature offset; Update the reference orientation field of the workpiece machining area based on the change in surface normal; The workpiece-level dynamic process reference constraint system is updated based on the updated dynamic process reference point set, dynamic process reference edge, dynamic process reference surface, and reference direction field.

5. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, In S3, the step of establishing the reference transfer chain includes: Based on the real-time pose data of the first heterogeneous robotic arm, establish the spatial transformation relationship between the base coordinate system of the first heterogeneous robotic arm and the end-effector coordinate system of the first heterogeneous robotic arm; Based on the workpiece-level dynamic process reference constraint system and workpiece spatial state and visual observation data, establish the spatial transformation relationship between the first heterogeneous robotic arm end coordinate system and the workpiece-level dynamic process reference constraint system. Based on the real-time pose data of the second heterogeneous robotic arm and the spatial state and visual observation data of the workpiece, a spatial transformation relationship is established between the workpiece-level dynamic process reference constraint system and the machining coordinate system of the second heterogeneous robotic arm. Based on the above spatial transformation relationship, a reference transfer chain is established from the workpiece-level dynamic process reference constraint system to the machining coordinate system of the second heterogeneous robotic arm.

6. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, In S3, the step of reconstructing the cooperative space mapping relationship includes: The relative spatial transformation relationship between the first heterogeneous robotic arm and the second heterogeneous robotic arm is calculated based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm. Based on the workpiece-level dynamic process reference constraint system and workpiece spatial state and visual observation data, the workpiece spatial constraint relationship is extracted. Establish a mapping relationship between the workpiece-level dynamic process datum constraint system and the visual coordinate system based on the workpiece-level dynamic process datum constraint system; Based on the workpiece-level dynamic process reference constraint system and the relative spatial transformation relationship between the first heterogeneous robotic arm and the second heterogeneous robotic arm, a mapping relationship is established from the workpiece-level dynamic process reference constraint system to the machining coordinate system of the second heterogeneous robotic arm. Based on the above mapping relationship and the relative spatial transformation relationship between the first heterogeneous robotic arm and the second heterogeneous robotic arm, a collaborative spatial mapping relationship is generated.

7. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, In S4, the step of updating the dynamic hand-eye mapping relationship between the workpiece-level dynamic process reference constraint system and the second heterogeneous robotic arm machining coordinate system includes: The offset of reference elements in the workpiece-level dynamic process datum constraint system is calculated based on the workpiece-level dynamic process datum constraint system and the collaborative space mapping relationship. The relative pose change is calculated based on the real-time pose data of the first heterogeneous robotic arm and the second heterogeneous robotic arm. The surface space deviation is calculated based on the collaborative spatial mapping relationship and the workpiece spatial state and visual observation data. The dynamic hand-eye mapping relationship is updated based on the offset, relative pose change, and surface space deviation of the reference elements in the workpiece-level dynamic process datum constraint system.

8. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, In S5, the step of correcting the processing trajectory of the second heterogeneous robotic arm based on the dynamic hand-eye mapping relationship includes: The trajectory offset of the workpiece machining area is calculated based on the dynamic hand-eye mapping relationship and the workpiece-level dynamic process datum constraint system. The workpiece machining area posture offset is calculated based on the workpiece spatial state, visual observation data, and dynamic hand-eye mapping relationship. The corrected machining trajectory of the second heterogeneous robotic arm is generated based on the trajectory offset and attitude offset. The current processing path of the second heterogeneous robotic arm is updated based on the corrected processing trajectory.

9. The method for large-scale spatial dynamic calibration and hand-eye error compensation of heterogeneous dual robotic arms according to claim 1, characterized in that, It also includes the following steps: Calculate the system state consistency index based on workpiece spatial state and visual observation data; Dynamic process baseline reliability parameters are generated based on system state consistency indicators. When the reliability parameter of the dynamic process reference is lower than the preset threshold, the reconstruction of the workpiece-level dynamic process reference constraint system, reference transfer chain and collaborative space mapping relationship is triggered.

Images