Intelligent identification and adaptive configuration method for robot end load

Abstract

The embodiment of the application discloses a kind of intelligent identification and adaptive configuration method of robot end load, it is related to robot control technical field.The method comprises: obtaining the cooperative handling trajectory when first robot and second robot jointly clamp long-size flexible workpiece;From the cooperative handling trajectory, determine the tentative motion section, and control first robot and second robot in the tentative motion section respectively execute in-phase impedance exploration and anti-phase impedance exploration;Obtain the stress change and pose change of the end of two robots during the in-phase impedance exploration;Obtain the force difference, relative displacement and response lag relationship of the end of two robots during the anti-phase impedance exploration;According to bending response characteristics, determine the flexible state of the long-size flexible workpiece;The parameters of first robot and second robot are configured respectively.The application improves the stress balance, motion stability and control safety in long-size flexible workpiece handling process.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method for intelligent identification and adaptive configuration of robot end effector loads. Background Technology

[0002] In the automated handling, assembly, and transfer of long workpieces, limitations imposed by the single robot's load capacity, workspace, and workpiece size typically necessitate the use of two robots, each gripping both ends of the workpiece and working together along a pre-set collaborative trajectory to complete the transport. To ensure transport stability, the workpiece's mass, center of gravity, and flexibility must be known in advance, and the load-bearing ratio, impedance control parameters, and movement speed of the two robots must be configured accordingly. However, in actual production, the workpiece's model, material, internal structure, and gripping position may change, causing discrepancies between the pre-entered load parameters and the actual conditions.

[0003] Existing load identification methods typically treat the workpiece to be transported as a rigid body, estimating its mass and center of gravity based on the force applied to the robot end effector, motion acceleration, or joint driving force. For long, flexible workpieces, the force difference generated by the two robot end effectors is not only related to the workpiece's center of gravity offset, but may also be caused by workpiece bending, local flexibility, hysteresis at both ends, and residual deformation.

[0004] Furthermore, existing dual-robot collaborative control typically employs fixed load-sharing ratios and impedance parameters, making it difficult to differentiate configurations based on the actual flexibility of the workpiece. When one side of the workpiece exhibits a stronger flexible response, fixed stiffness control may cause that side to bear excessive internal forces, resulting in increased workpiece deflection, slippage in the clamping position, or accumulation of collaborative errors between the two robots. Therefore, it is necessary to provide a method capable of distinguishing between the rigid motion and bending response of the workpiece and adaptively configuring the control parameters of the two robots based on the identification results. Summary of the Invention

[0005] This application provides a method for intelligent identification and adaptive configuration of robot end-effector load, which improves the force balance, motion stability and control safety during the handling of long flexible workpieces.

[0006] This application provides the following solution: According to a first aspect, a method for intelligent identification and adaptive configuration of robot end-effector loads is provided, comprising: acquiring a collaborative transport trajectory when a first robot and a second robot jointly clamp a long flexible workpiece; determining a trial motion segment from the collaborative transport trajectory, and controlling the first robot and the second robot to perform in-phase impedance testing and out-of-phase impedance testing respectively in the trial motion segment; wherein, the in-phase impedance testing refers to applying virtual impedance changes in the same direction and time synchronously to the first robot and the second robot, causing the end-effectors of the two robots to produce pose changes in the same direction; the out-of-phase impedance testing refers to applying virtual impedance changes in opposite directions to the first robot and the second robot, causing the end-effectors of the two robots to produce relative displacement; The system acquires the force and pose changes of the two robot ends during the in-phase impedance test, and generates rigid body motion response characteristics of the long flexible workpiece based on the consistent relationship between the force and pose changes. It also acquires the force difference, relative displacement, and response hysteresis relationship of the two robot ends during the out-of-phase impedance test, and generates bending response characteristics of the long flexible workpiece. Based on the rigid body motion response characteristics, it determines the mass and center of gravity position of the long flexible workpiece, and determines its flexibility state based on the bending response characteristics. Finally, based on the mass, center of gravity position, and flexibility state, it configures the load sharing ratio, virtual stiffness parameters, virtual damping parameters, and motion restriction parameters for the first and second robots, respectively.

[0007] According to one achievable method in the embodiments of this application, determining the trial motion segment from the collaborative transport trajectory includes: dividing the collaborative transport trajectory into multiple candidate motion segments; and determining the trial motion segment that meets the preset trial conditions based on the motion acceleration change corresponding to each candidate motion segment, the force margin of the two robot ends, and the workpiece pose deviation margin.

[0008] According to one achievable method in an embodiment of this application, the in-phase impedance test includes: applying virtual impedance changes in the same direction and time to the first robot and the second robot, so that the ends of the two robots produce small pose changes in the same direction, while keeping the predetermined transport direction of the long flexible workpiece unchanged.

[0009] According to one achievable method in the embodiments of this application, the reverse impedance probing includes: applying virtual impedance changes in opposite directions to the first robot and the second robot to generate relative displacement at the ends of the two robots, and adjusting the motion amplitude of the ends of the two robots according to the relative displacement to limit the overall positional offset of the long flexible workpiece.

[0010] According to one achievable method in the embodiments of this application, generating the rigid body motion response features of the long flexible workpiece based on the consistent relationship between force changes and pose changes includes: extracting the same-direction component in the force changes of the two robot ends and the synchronous component in the pose changes; and generating the rigid body motion response features based on the same-direction component, the synchronous component, and the response phase relationship between the two.

[0011] According to one achievable method in an embodiment of this application, generating the bending response characteristics of the long-sized flexible workpiece includes: determining the force difference based on the force changes at the ends of the first and second robots, and determining the relative displacement based on the pose changes at the ends of the two robots; determining the response hysteresis relationship between the moment when the force difference reaches its peak and the moment when the relative displacement reaches its peak; and generating the bending response characteristics based on the magnitude of the force difference, the magnitude of the relative displacement, and the response hysteresis relationship.

[0012] According to one achievable method in an embodiment of this application, determining the mass and center of gravity position of the long flexible workpiece based on the rigid body motion response characteristics includes: determining multiple effective rigid body response time periods based on the consistency between the force changes and pose changes of the two robot ends in the rigid body motion response characteristics; generating multiple candidate masses based on the resultant force and common acceleration of the two robot ends within each effective rigid body response time period; generating corresponding candidate center of gravity positions based on each candidate mass, the force distribution relationship of the two robot ends, and the gripping position; determining the concentrated distribution range of the multiple candidate masses and candidate center of gravity positions, and determining the mass and center of gravity position of the long flexible workpiece based on the concentrated distribution range.

[0013] According to one achievable method in an embodiment of this application, after generating the corresponding candidate center of gravity position, the method further includes: arranging the candidate center of gravity positions in chronological order according to the effective time period of each rigid body response, and determining the migration direction of the candidate center of gravity positions; matching the migration direction with the relative displacement direction and response hysteresis relationship corresponding to the bending response feature; marking the candidate center of gravity positions whose migration direction changes synchronously with the bending response as deformation interference positions; performing reverse compensation on the deformation interference positions according to the relative displacement change amount of the corresponding time period, and determining the center of gravity position of the long flexible workpiece based on the compensated candidate center of gravity positions.

[0014] According to one achievable method in an embodiment of this application, determining the flexibility state of the long flexible workpiece based on the bending response characteristics includes: determining a bending loading stage and a bending recovery stage based on the increasing and decreasing processes of the force difference in the bending response characteristics; comparing the relative displacements corresponding to the bending loading stage and the bending recovery stage under the same force difference to determine the degree of bending hysteresis; determining the remaining relative displacement after the force difference recovers to the reference range, and determining the dominant side of the flexible response based on the response hysteresis relationship; and determining an elastic bending state, a lateral flexible state, or a continuous deformation state based on the degree of bending hysteresis, the remaining relative displacement, and the dominant side of the flexible response.

[0015] According to one achievable method in an embodiment of this application, configuring the load-sharing ratio, virtual stiffness parameter, and virtual damping parameter of the first robot and the second robot based on the mass, center of gravity position, and flexibility state includes: determining the basic load-sharing ratio of the two robots based on the center of gravity position; determining the robot to be unloaded and the allowable load migration amount based on the dominant side of the flexibility response, the degree of bending hysteresis, and the remaining relative displacement; reducing the virtual stiffness of the robot to be unloaded and increasing the virtual stiffness of the other robot while keeping the combined support force of the two robots within a preset range, so that the allowable load migration amount is gradually transferred to the other robot; adjusting the virtual damping parameters of the two robots based on the force difference and relative displacement change during the load migration process, and determining the load-sharing ratio when the force difference and relative displacement are restored to the corresponding allowable range as the target load-sharing ratio.

[0016] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application employs in-phase and out-of-phase impedance testing during the collaborative handling of long, flexible workpieces by dual robots. This allows for the extraction of rigid body motion response characteristics corresponding to the overall motion of the workpiece and bending response characteristics corresponding to the relative deformation at both ends. This effectively distinguishes between load changes caused by mass and center of gravity and force differences caused by flexible bending. Determining the workpiece's mass and center of gravity based on rigid body motion response characteristics, and combining this with bending response characteristics to identify the workpiece's flexible state, reduces the interference of local bending, response hysteresis, and residual deformation on load identification results. Furthermore, adaptively configuring the load-sharing ratio, virtual stiffness, virtual damping, and motion limitation parameters of the two robots based on the identification results helps reduce collaborative internal forces and workpiece deflection, improving force balance, motion stability, and control safety during the handling of long, flexible workpieces.

[0017] Of course, any product implementing this application does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating the intelligent identification and adaptive configuration method for robot end effector loads provided in this application embodiment; Figure 2 This is a schematic diagram of a structure provided in an embodiment of this application for a dual-robot joint clamping of a long, flexible workpiece; Figure 3 This is a schematic diagram illustrating the collaborative transport trajectory division and trial motion segment determination provided in the embodiments of this application; Figure 4 This is a schematic diagram showing the mass, candidate center of gravity position, and deformation interference compensation provided for embodiments of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0021] Figure 1 A flowchart illustrating the intelligent identification and adaptive configuration method for robot end effector loads provided in this application embodiment. Figure 1 As shown, the method may include the following steps: Step 101: Obtain the collaborative transport trajectory when the first robot and the second robot jointly clamp the long flexible workpiece.

[0022] Step 102: Determine the trial motion section from the cooperative transport trajectory, and control the first robot and the second robot to perform in-phase impedance testing and out-of-phase impedance testing respectively in the trial motion section.

[0023] Step 103: Obtain the force changes and pose changes of the two robot ends during the in-phase impedance test, and generate the rigid body motion response characteristics of the long flexible workpiece based on the consistent relationship between the force changes and pose changes.

[0024] Step 104: Obtain the force difference, relative displacement and response hysteresis relationship of the two robot ends during the reverse impedance test, and generate the bending response characteristics of the long flexible workpiece.

[0025] Step 105: Determine the mass and center of gravity of the long flexible workpiece based on the rigid body motion response characteristics, and determine the flexibility state of the long flexible workpiece based on the bending response characteristics.

[0026] Step 106: Based on the mass, center of gravity position, and flexibility state, configure the load sharing ratio, virtual stiffness parameters, virtual damping parameters, and motion restriction parameters of the first robot and the second robot respectively.

[0027] As can be seen from the above process, this application, by setting up in-phase impedance testing and out-of-phase impedance testing during the collaborative handling of long flexible workpieces by two robots, extracts the rigid body motion response characteristics corresponding to the overall motion of the workpiece and the bending response characteristics corresponding to the relative deformation at both ends, thereby effectively distinguishing the load changes caused by mass and center of gravity from the force differences caused by flexible bending. Determining the workpiece mass and center of gravity position based on the rigid body motion response characteristics, and combining this with the bending response characteristics to identify the workpiece's flexible state, reduces the interference of local bending, response hysteresis, and residual deformation on the load identification results. Furthermore, adaptively configuring the load sharing ratio, virtual stiffness, virtual damping, and motion limitation parameters of the two robots based on the identification results helps reduce collaborative internal forces and workpiece deflection, improving the force balance, motion stability, and control safety during the handling of long flexible workpieces.

[0028] The following describes in detail each step of the above process and the effects that can be further produced, with reference to the embodiments.

[0029] Step 101 specifically involves obtaining the collaborative transport trajectory when the first robot and the second robot jointly clamp a long flexible workpiece.

[0030] Figure 2 This is a schematic diagram of a structure provided in this application embodiment for dual robots jointly clamping a long, flexible workpiece. Obtaining the collaborative transport trajectory when the first and second robots jointly clamp the long, flexible workpiece refers to acquiring the timing motion information used to control the two robots to jointly complete lifting, translation, turning, deceleration, and placement actions after the two robots respectively clamp the workpiece at the first and second gripping positions. The collaborative transport trajectory can be pre-planned and generated by the host controller based on the initial position of the workpiece, the target placement position, the workspace of the two robots, and the distribution of obstacles, or it can be obtained in real-time by the robot controller based on preset path points and motion constraints during the execution of the transport task.

[0031] The collaborative handling trajectory can include the changes in the workpiece reference position and the workpiece reference orientation over time, as well as the corresponding motion velocity and acceleration. Based on the workpiece reference position and orientation, and combined with the fixed relationship between the first and second gripping positions and the workpiece reference coordinate system, the end-effector target trajectories of the first and second robots are calculated respectively. This allows the two robot end-effectors to jointly move the long, flexible workpiece while maintaining a predetermined gripping distance and relative orientation, avoiding significant motion asynchrony and additional internal forces caused by the two robots independently planning their trajectories.

[0032] In one specific implementation, a workpiece reference coordinate system is established using the geometric center of the long, flexible workpiece or the midpoint of the line connecting the two gripping positions. The end-effector poses of the first and second robots are uniformly transformed to this workpiece reference coordinate system. The collaborative transport trajectory records the workpiece target pose, the first robot end-effector target pose, and the second robot end-effector target pose at each sampling time according to a unified time reference. If the sampling periods of the two robot controllers differ or there is a time deviation in communication, the trajectory data can be aligned based on timestamps, and missing trajectory sampling points can be supplemented through interpolation to ensure the comparability of the motion states of the two robots at the same sampling time.

[0033] The cooperative transport trajectory can be either a planned trajectory not yet executed by the controller or the actual trajectory recorded during robot execution. The planned trajectory is used to determine candidate motion segments for subsequent impedance testing, while the actual trajectory is used to determine whether the two robots move according to the predetermined cooperative relationship. When the deviation between the actual trajectory and the planned trajectory exceeds the allowable range, or when the relative position change between the two robot ends exceeds the preset range, impedance testing can be temporarily suspended to avoid introducing additional pose changes when the cooperative state is unstable.

[0034] Step 102 specifically involves: determining the trial motion segment from the cooperative transport trajectory, and controlling the first robot and the second robot to perform in-phase impedance testing and out-of-phase impedance testing respectively in the trial motion segment.

[0035] Determining the test motion segment from the collaborative handling trajectory refers to selecting suitable trajectory segments from the complete handling process, without altering the normal handling task of long, flexible workpieces or reducing clamping safety, to perform in-phase and out-of-phase impedance testing. Because the force states and pose stability vary during the lifting, acceleration, constant speed, turning, deceleration, and placement phases, not all trajectory segments are suitable for impedance testing. Therefore, it is necessary to first segment the collaborative handling trajectory and then select segments based on their motion states and safety margins.

[0036] Figure 2This is a schematic diagram of a structure provided in this application embodiment for dual robots jointly clamping a long, flexible workpiece. Specifically, the collaborative handling trajectory can be divided into multiple candidate motion segments according to changes in motion speed, acceleration, direction, and workpiece posture. When the motion state remains continuous at adjacent sampling times, and the motion direction and motion control mode do not change significantly, the corresponding trajectory data are assigned to the same candidate motion segment. When it is detected that the robot enters a lifting state from a stationary state, a constant speed state from an accelerating state, a change in motion direction, an adjustment of the workpiece posture, or a deceleration state, the corresponding time is used as the boundary point of the candidate motion segment.

[0037] For each candidate motion segment, the motion acceleration corresponding to the workpiece reference position or the common pose of the two robot end effectors within that segment is obtained, and the amplitude and continuity of the motion acceleration variation are determined. When the motion acceleration variation is too large, the force on the robot end effector includes a strong inertial impact, which can easily mask the subtle response caused by impedance probing. When the motion acceleration changes frequently, it is also difficult to accurately distinguish between the normal trajectory response and the probing response. Therefore, candidate motion segments with relatively stable motion acceleration or continuous variation in a single direction can be prioritized, while candidate motion segments corresponding to abrupt starts, emergency decelerations, rapid turns, and sudden attitude changes can be excluded.

[0038] The end-effector force margins for the two robots are determined based on the differences between the current end-effector forces of the first and second robots and their corresponding upper limits of allowable force. End-effector forces can include forces acting along different coordinate axes and moments about different coordinate axes. For each candidate motion segment, the remaining allowable forces in each force direction for both robots can be calculated separately, and the smaller remaining allowable force is taken as the end-effector force margin for that robot in the current segment. When the end-effector force of either robot is close to the upper limit of allowable force, impedance testing is not performed in that segment to avoid overloading the robot, loosening the gripper, or causing excessive localized stress on the workpiece due to the superposition of the testing effect and the normal handling load.

[0039] The workpiece pose deviation margin is determined based on the current deviation between the actual workpiece pose and the target pose in the collaborative transport trajectory, as well as the maximum allowable pose deviation of the system. Specifically, the workpiece's position deviation, attitude deviation, and the relative position deviation between the two robot end effectors can be obtained, and the remaining amount corresponding to the allowable upper limit for each deviation distance can be determined. When the workpiece's current pose is close to the allowable deviation boundary, it indicates insufficient collaborative motion stability, and it is not advisable to continue adding trial displacements. When the workpiece pose deviation is small, and the relative position between the two robot end effectors remains stable, it indicates that this section has a large pose deviation margin, allowing the two robots to perform small impedance tests with limited amplitude.

[0040] The preset trial conditions may include motion acceleration changes below a corresponding change threshold, end-effector force margins of both the first and second robots exceeding force margin thresholds, and workpiece pose deviation margins exceeding pose margin thresholds. Only candidate motion segments that simultaneously meet the above conditions are determined as usable trial motion segments. When multiple candidate motion segments meet the preset trial conditions, the duration, motion stability, and force margin of each segment can be further compared, prioritizing segments with longer durations, smaller motion acceleration changes, and larger force margins for both robots.

[0041] Controlling the first and second robots to perform in-phase impedance testing and out-of-phase impedance testing respectively within the testing motion segment means that while the two robots continue to move the long, flexible workpiece along the cooperative transport trajectory, amplitude-limited testing changes are sequentially superimposed onto the original impedance control parameters or end-effector reference poses of the two robots. This causes the workpiece to produce a response dominated by overall motion and a response dominated by relative bending, respectively. The two types of impedance testing can be performed sequentially within the same testing motion segment, or separately within two testing motion segments that meet preset testing conditions. When using the same testing motion segment, a recovery period can be set between the two types of testing to allow the end-effector forces and workpiece poses of the two robots to return to the reference range before the testing, avoiding the residual response from the previous test from affecting the subsequent test.

[0042] Virtual impedance change refers to a control quantity that temporarily adjusts the impedance control relationship of the robot's end effector. It can manifest as virtual stiffness change, virtual damping change, or a small change in reference pose associated with virtual impedance control. The robot controller superimposes the virtual impedance change onto the normal cooperative handling control command, causing the robot's end effector to produce detectable small force and pose changes without deviating from the original cooperative handling trajectory. The direction of the virtual impedance change can be determined based on the workpiece handling direction, workpiece length direction, gravity direction, or predetermined bending detection direction, and its amplitude should be lower than the limit value that would cause fixture slippage, permanent workpiece deformation, or significant deviation from the cooperative trajectory.

[0043] During in-phase impedance testing, the first and second robots receive virtual impedance changes in the same direction and at the same start time. For example, both robot end effectors may superimpose a small reference displacement in the same direction along the current workpiece transport direction, or synchronously reduce the virtual stiffness in the corresponding direction, causing the two robot end effectors to produce small pose changes in the same direction under external force. Time synchronization means that the start time, duration, and rate of change of the two virtual impedance changes are consistent, allowing for a small amount of time compensation to be pre-set based on the response differences between the two robot controllers, so that the actual end effector responses are basically synchronized.

[0044] When the two robot end effectors move in the same direction, the relative distance and relative orientation between the first and second gripping positions remain essentially unchanged, and long, flexible workpieces mainly exhibit overall translation or rotation. In this situation, the force changes on the two robot end effectors typically have the same direction of change, and the end effector pose changes also exhibit high synchronicity. Therefore, the response generated by in-phase impedance probing can be used to extract the rigid body motion response of the workpiece, thus providing a data basis for mass and center of gravity position identification.

[0045] Maintaining the predetermined transport direction of a long, flexible workpiece means that minor pose changes caused by in-phase impedance probing do not alter the primary direction of motion specified in the cooperative transport trajectory. In practice, the in-phase probing direction can be set to be consistent with the current transport direction, or the probing changes can be limited to permissible directions that do not alter the overall orientation of the workpiece. While superimposing the probing changes, the robot controller continues to execute the original cooperative transport speed and attitude commands, ensuring that both robot end effectors smoothly return to the original cooperative transport trajectory after the probing is completed, avoiding path deviation or interruption of the transport cycle during the probing process.

[0046] During the reverse impedance probing process, virtual impedance changes in opposite directions are applied to the first and second robots. For example, the end effector of the first robot generates a small positive displacement along a predetermined probing direction, while the end effector of the second robot generates a small reverse displacement along the same axis. Alternatively, the virtual stiffness of the first robot can be reduced along this direction, while the second robot performs a small reference pose adjustment in the opposite direction. By moving in opposite directions at both ends, a controlled relative displacement is generated between the first and second gripping positions, thereby creating a small bending effect on the long, flexible workpiece.

[0047] During the reverse impedance test, the relative displacement of the two robot ends is primarily used to elicit the workpiece's flexible response. For workpieces with low flexibility, the force difference may increase rapidly, while the relative displacement is relatively small. For workpieces with high flexibility, a large relative displacement may occur under the same force difference, potentially leading to significant response hysteresis or slow recovery. By collecting the force difference, relative displacement, and their time relationship between the two robot ends during the reverse impedance test, the bending capacity and recovery characteristics of different workpieces can be distinguished.

[0048] To prevent the reverse probing from causing a shift in the overall workpiece position, the motion amplitudes of the two robot end effectors need to be adjusted based on the real-time relative displacement. Specifically, the displacements of the first and second robot end effectors can be converted to a unified workpiece reference coordinate system, and the common and differential components of the two robot end effector displacements can be calculated. The common component characterizes the overall workpiece position shift, and the differential component characterizes the relative displacement between the two gripping positions. When the common component increases, the probing amplitude of the robot on the side producing the larger displacement is reduced, or the reverse probing amplitude of the other robot is increased, so that the effects of the displacements at both ends on the overall workpiece position cancel each other out.

[0049] In one specific implementation, the displacement amplitudes of the two robot end effectors along the probing direction can be initially set to be the same. When the local flexibility between the first robot and the workpiece is relatively large, causing the actual displacement of the first robot end effector to be greater than that of the second robot end effector, the controller reduces the subsequent probing amplitude of the first robot and appropriately increases the reverse probing amplitude of the second robot. After adjustment, the average displacement of the two robot end effectors is kept within the allowable range, while maintaining sufficient relative displacement to stimulate the workpiece bending response. This limits significant offset of the geometric center or the midpoint between the two gripping positions of a long, flexible workpiece.

[0050] Both in-phase and out-of-phase impedance tests can apply virtual impedance changes by gradually increasing and then gradually decreasing them to avoid strong impacts from abrupt changes. During the test, the force on the end effectors of both robots, their relative displacement, the overall workpiece pose, and the clamping status are continuously monitored. If the end effector force exceeds the allowable range, the relative displacement exceeds the deformation limit, the overall workpiece position deviation exceeds the allowable value, or the clamping status is abnormal, the test should be stopped immediately and the original virtual impedance parameters restored.

[0051] Step 103 specifically involves: obtaining the force and pose changes of the two robot ends during the in-phase impedance test, and generating the rigid body motion response characteristics of the long-sized flexible workpiece based on the consistent relationship between the force and pose changes.

[0052] Acquiring the force and pose changes of the two robot end effectors during in-phase impedance testing refers to continuously collecting force, torque, position, and attitude data of the two robot end effectors according to a unified time reference after the first and second robots receive virtual impedance changes in the same direction and time synchronously, and extracting the changes relative to the reference state before the start of the in-phase impedance test. The force changes can be obtained from force sensors installed on the end effectors of the two robots, or calculated based on joint drive torque, robot dynamics model, and end effector Jacobian relationships. The pose changes can be determined by the end effector position and attitude fed back by the robot controller.

[0053] Before data acquisition, a settling period before the start of the in-phase impedance test can be selected as a reference period. The average force and average pose of the first and second robot end effectors during this reference period are calculated. The force at each sampling moment during the test is subtracted from the corresponding average force to obtain the force changes of the two robot end effectors. The pose of the end effectors at each sampling moment during the test is compared with the corresponding average pose to obtain the position and attitude changes of the two robot end effectors. By using relative changes, the influence of workpiece weight, clamping force, and sensor zero-point offset on the test response analysis can be reduced.

[0054] Since the first and second robots may be located in different installation positions, and their end effector sensors may use different coordinate directions, it is necessary to transform the force and pose changes of the two robots' end effectors to a unified workpiece reference coordinate system. The workpiece reference coordinate system can be established at the midpoint between the two gripping positions, with one coordinate axis set along the workpiece length direction and the other along the in-phase impedance probe direction. After coordinate transformation, force and pose changes along the same coordinate axis have consistent physical meanings and can be used to determine whether the two robots jointly drive the workpiece to produce overall motion.

[0055] The consistency between force changes and pose changes is mainly used to characterize whether the responses of two robot end effects have the same direction of change, similar timing of change, and continuous trend. Specifically, the positive and negative directions of force changes at the two robot end effects can be compared. When the force changes at the two robot end effects along the same coordinate axis increase or decrease simultaneously, they are considered to have a force relationship in the same direction. The direction and timing of pose changes at the two robot end effects can also be compared. When the two robot end effects generate displacement along the same direction, and the start time and peak time of the displacement are close, they are considered to have a synchronous pose relationship.

[0056] In one specific implementation, a co-directional component can be extracted from the force changes at the end effectors of the first and second robots. This co-directional component characterizes the combined load change caused by the combined action of the two robots on a long, flexible workpiece. For portions where the forces at the end effectors of the two robots are in opposite directions, this can be considered a differential response caused by internal forces or local deformations, and not used as the primary rigid body response. Similarly, a synchronization component can be extracted from the position and attitude changes of the two robot end effectors to characterize the translation or rotation occurring simultaneously at the two gripping positions.

[0057] When extracting the synchronization component, the position changes of the first and second robot end effectors at the same sampling time can be combined to obtain the common displacement of the two gripping positions. The overall rotational change of the workpiece can also be determined based on the common direction of the attitude changes of the two robot end effectors. When the relative distance between the two gripping positions remains essentially constant, and the common displacement of the two robot end effectors is significantly greater than the relative displacement, it can be assumed that the long, flexible workpiece mainly undergoes rigid body translation during this period. When the attitude changes of the two robot end effectors are in the same direction, and the two gripping positions rotate around the same direction, it can be assumed that the workpiece mainly undergoes rigid body rotation.

[0058] Furthermore, the response phase relationship between the co-directional force component and the synchronous pose component can be determined. This response phase relationship characterizes the degree of time correspondence between the change in end-effector force and the overall motion change of the workpiece. For example, the moments when the co-directional force component begins to change and reaches its peak value, as well as the moments when the synchronous pose component begins to change and reaches its peak value, can be determined, and the response phase relationship can be determined based on the time difference between these corresponding moments. When the time difference between the force change and the pose change is within an allowable range, it indicates that both are jointly triggered by the same in-phase impedance test, and can be considered as an effective rigid body motion response.

[0059] When generating rigid body motion response characteristics, the degree of simultaneity of force changes at the ends of the two robots, the degree of synchronization of pose changes, the amplitude of changes in simultaneous force components, the amplitude of changes in synchronized pose components, and the phase relationship between the two responses can be combined. These rigid body motion response characteristics can be represented using a time-series feature set, feature vector, or associated data structure, and stored in the temporal order of in-phase impedance probes. This feature not only reflects the combined force required for the overall motion of a long, flexible workpiece, but also the response speed and stability of the workpiece's overall motion to the probe input.

[0060] When the forces and poses of the two robot ends change in the same direction, and there is a stable temporal correspondence between the force and pose changes, the rigid body motion response during that time period can be considered relatively significant. When the forces on the two robot ends change in different directions, or when the relative displacement between the ends increases significantly, it indicates that the time period may be affected by workpiece bending, clamping slippage, or cooperative errors. This can reduce the effectiveness of that time period in the rigid body motion response characteristics, or it may not be used for subsequent mass and center of gravity position identification.

[0061] Through the above processing, the response part of the workpiece motion jointly driven by the two robots can be extracted from the mixed response generated by the in-phase impedance test, so that the generated rigid body motion response characteristics mainly reflect the overall inertia and overall load distribution of the long flexible workpiece, reducing the interference of local bending and internal forces at both ends on the subsequent mass and center of gravity position calculation.

[0062] Step 104 specifically involves: obtaining the force difference, relative displacement, and response hysteresis relationship of the two robot ends during the anti-phase impedance test, and generating the bending response characteristics of the long-sized flexible workpiece.

[0063] Acquiring the force difference, relative displacement, and response hysteresis relationship of the two robot end effectors during the reverse impedance test involves continuously collecting force and pose data from the end effectors of the first and second robots after they receive virtual impedance changes in opposite directions. The bending deformation information of the long, flexible workpiece is then extracted based on the difference in the responses of the two end effectors. Because the reverse impedance test causes controlled motion in opposite directions at the two gripping positions, the overall translation of the workpiece is restricted. The collected differential responses at both ends primarily reflect the bending transmission and deformation recovery characteristics of the workpiece between the two gripping positions.

[0064] Before the reverse impedance test begins, a period of relatively stable force and pose can be selected as a reference time period to determine the reference force and pose of the first and second robot end effectors. The end effector force at each sampling moment during the reverse impedance test is compared with the corresponding reference force to obtain the force changes of the first and second robot end effectors. The end effector pose at each sampling moment is compared with the corresponding reference pose to obtain the position and attitude changes of the two robot end effectors. By extracting the changes relative to the reference state, the influence of workpiece weight, clamping force, and initial sensor offset on the bending response analysis can be reduced.

[0065] The force data of the first and second robots can be collected by force sensors installed between the robot end effectors and the gripping mechanism, or calculated based on the robot joint drive torques and robot dynamics models. The force data can include end effector forces and torques. Before processing, the force changes at the ends of the two robots are converted to a unified workpiece reference coordinate system, and the positive directions of the two end effector force data are unified, allowing for direct comparison of the effects of the two robots in the predetermined bending direction.

[0066] When determining the force difference based on the force changes at the ends of the first and second robots, the difference in force changes at the two ends in the predetermined bending direction at the same sampling moment can be calculated. Since the two robots apply forces in opposite directions during the reverse impedance test, the degree of imbalance of the forces at both ends can be determined after unifying the force direction. A large force difference indicates that the bending load transmitted between the two gripping positions of the workpiece is large, or that one side has a higher resistance to the test input. A small force difference indicates that the workpiece can easily release the reverse forces at both ends through deformation, or that the force responses at both ends are relatively similar.

[0067] In one implementation, the starting moment of the increase in force difference, the maximum value, the time required to reach the maximum value, the decrease process, and the time required to recover to the reference range can be recorded separately. The magnitude of the force difference change can be represented by the difference between the maximum force difference and the reference force difference before the test, or it can be determined based on the cumulative change in force difference during the entire test period. If the test direction changes, the positive and negative directions of the force difference can also be retained to characterize whether the bending effect of the workpiece is biased towards the first robot side or the second robot side.

[0068] When determining relative displacement based on the pose changes of the two robot end effectors, the position changes of the first and second robot end effectors can be converted to the workpiece reference coordinate system, and the displacement difference between the two in the predetermined bending direction can be calculated. This displacement difference reflects the degree of relative movement between the two gripping positions. The relative rotation between the two gripping positions can also be determined based on the difference in the posture changes of the two robot end effectors. For long, flexible workpieces that primarily exhibit lateral deflection, the difference in lateral displacement between the two ends can be used as the relative displacement. For workpieces exhibiting both deflection and torsion, both relative position changes and relative posture changes can be used as relative displacement information.

[0069] The magnitude of the relative displacement change can be determined based on the difference between the maximum relative displacement during the reverse impedance test and the baseline relative displacement value before the test. To reduce the influence of the overall workpiece position offset, the common displacement and differential displacement can be separated from the displacements of the two robot end effectors. The average change in the displacements of the two robot end effectors is used to characterize the overall workpiece position offset, while the difference between the two displacements is used to characterize the relative displacement caused by workpiece bending. The differential displacement is primarily used when generating bending response characteristics subsequently.

[0070] When determining the response hysteresis relationship, peak values ​​are identified in both the force difference change sequence and the relative displacement change sequence. The peak moment of the force difference refers to the moment when the force difference reaches its maximum change during the reverse impedance test. The peak moment of the relative displacement refers to the moment when the relative displacement between the two robot ends reaches its maximum change. The time difference between the peak moment of the relative displacement and the peak moment of the force difference is calculated, and this time difference is used as a representation of the response hysteresis relationship.

[0071] When the force difference reaches its peak first, and the relative displacement reaches its peak after a period of time, it indicates that the opposing forces first form at both ends of the workpiece, and the bending deformation then gradually propagates towards the middle of the workpiece. A larger time difference between the peak values ​​usually indicates a slower establishment speed of the workpiece's bending response, or a significant hysteresis in the flexible transmission within the workpiece. When the peak values ​​of the force difference and the relative displacement are essentially synchronized, it indicates that the workpiece can quickly form the corresponding deformation after being subjected to opposing forces from both ends. When the relative displacement remains at a high level even after the force difference has decreased, it can also indicate that the workpiece has slow recovery or residual deformation.

[0072] When sensor noise or multiple local peaks exist near the peak value, the force difference sequence and relative displacement sequence can be smoothed first, and then the target peak value can be determined from the peak values ​​whose continuous holding time and change amplitude both meet the preset requirements. Alternatively, a time window can be set near the target peak value, and the peak time can be determined based on the concentrated change position of each sample value within the time window, so as to avoid errors in the calculation of response lag relationship caused by a single abnormal sample value.

[0073] When generating bending response characteristics based on the variation range of force difference, the variation range of relative displacement, and the response hysteresis relationship, these three factors can be associated and stored according to the time sequence of the same anti-phase impedance test process. The bending response characteristics may include the variation range of force difference, the variation range of relative displacement, the peak time of force difference, the peak time of relative displacement, the peak time difference, and the corresponding test direction. Bending response characteristics can be represented using feature vectors, time-series feature sets, or structured parameter sets.

[0074] In one specific implementation, a large change in force difference, a small change in relative displacement, and a short response lag time indicate that the workpiece has strong resistance to bending during reverse testing. A small change in force difference and a large change in relative displacement indicate that the workpiece is prone to bending. When both the change in force difference and the change in relative displacement are large, and the relative displacement significantly lags behind the force difference, it indicates that the transmission and release of bending load within the workpiece are slow. The aforementioned response states are further used in subsequent steps, combining the loading stage, the recovery stage, and the remaining relative displacement, to determine the workpiece's flexibility.

[0075] Through the above processing, a controlled two-end difference response can be actively formed by using reverse impedance probing, and the correlation between force, deformation and time lag can be extracted from it. This allows the generated bending response characteristics to reflect the bending degree, bending response speed and two-end flexibility difference of long flexible workpieces, providing a basis for subsequent identification of elastic bending state, lateral flexibility state or continuous deformation state.

[0076] Step 105 specifically involves: determining the mass and center of gravity of the long flexible workpiece based on the rigid body motion response characteristics, and determining the flexibility state of the long flexible workpiece based on the bending response characteristics.

[0077] Determining the mass and center of gravity of a long flexible workpiece based on the rigid body motion response characteristics means selecting effective data that is less affected by flexible bending from the joint motion response of the two robots generated by in-phase impedance testing, and estimating the workpiece mass and center of gravity based on the resultant force applied to the workpiece by the two robots, the joint motion acceleration of the workpiece, and the force distribution at both ends. Figure 4 This is a schematic diagram showing the mass, candidate center of gravity position, and deformation interference compensation provided for embodiments of this application.

[0078] Since a single sampling moment is easily affected by sensor noise, control delay and local vibration, this implementation does not directly use the calculation result of a single moment. Instead, it generates candidate results in multiple effective rigid body response periods and then determines the final identification result based on the concentrated distribution of the candidate results.

[0079] Specifically, the effective time period of rigid body response is determined based on the consistency between the force changes and pose changes at the ends of the two robots in the rigid body motion response characteristics. This can be achieved by comparing the consistency of the direction of force changes and pose changes at the ends of the first and second robots, as well as the temporal correspondence between the force changes and pose changes. When the directions of force changes and pose changes are the same for both robots, and the relative displacement between the ends remains within the allowable range, it indicates that the workpiece mainly undergoes overall motion, and the corresponding continuous time range can be determined as the effective time period of rigid body response. When the directions of force changes at the ends are opposite, the relative displacement increases significantly, or the response time differs greatly, it indicates that this time period may contain bending response or cooperative error, and it is not considered as the effective time period of rigid body response.

[0080] After determining multiple effective time periods for rigid body response, the forces acting on the first and second robot end effectors along their common direction of motion are obtained within each effective time period, and these two forces are converted to a unified workpiece reference coordinate system. The two end effector forces are then combined to obtain the resultant force of the two robots acting together on the long, flexible workpiece. To reduce the influence of the robot end effector's own weight, the mass of the clamping mechanism, and the fixing support force, the robot body's force reference when the workpiece is not clamped can be obtained in advance, and the corresponding reference component can be subtracted from the actual resultant force.

[0081] The common motion acceleration can be determined based on the synchronization component in the pose changes of the two robot end effectors. Specifically, the motion velocity and acceleration of the first and second robots are calculated separately based on their respective end effector position changes. Then, the portions of change with the same direction and synchronized time are extracted to form the common motion acceleration of the workpiece. Alternatively, the common motion acceleration can be calculated based on the common displacement of the two gripping positions or the displacement change of their midpoints. For cases where there is simultaneous overall rotation, the common angular acceleration of the workpiece can also be obtained, and the effective time period with a more obvious overall translational response is prioritized during quality identification.

[0082] For each effective time period of the rigid body response, candidate masses are generated based on the resultant force and common acceleration within that time period. In practice, multiple sampling points can be selected within the effective time period, and the instantaneous mass can be calculated based on the correspondence between the resultant force and the common acceleration. Then, the multiple instantaneous masses are averaged or fitted to obtain the candidate mass corresponding to that effective time period. When the transport direction includes a gravity component, the known stable component in the gravity direction can be subtracted from the resultant force, or a trial response in the horizontal direction can be selected to calculate the candidate mass, in order to avoid calculation errors caused by the mixture of workpiece weight and inertia.

[0083] When generating candidate center of gravity positions based on candidate masses, the force distribution relationship between the two robot end effectors, and the gripping positions, the first step is to determine the load share borne by the first and second robots during the corresponding rigid body response effective time periods. The force distribution relationship can be determined based on the proportion of the force exerted by the two robot end effectors along the support direction to the resultant force. The gripping positions can be represented by the coordinates of the first and second gripping positions in the workpiece reference coordinate system. Based on the total load corresponding to the candidate masses and the load shares at both ends, and according to the moment balance relationship between the two gripping positions and the center of gravity, the candidate position of the workpiece center of gravity along the line connecting the two gripping positions is determined.

[0084] For example, when the first robot bears a larger share of the load than the second robot, the workpiece's center of gravity can be preliminarily determined to be biased towards the first gripping position. When the second robot bears a larger share of the load, the workpiece's center of gravity can be determined to be biased towards the second gripping position. The more significant the difference in force distribution, the greater the degree to which the candidate center of gravity position shifts towards the side bearing the larger load. For three-dimensional material handling scenarios, the projected positions of the center of gravity in multiple directions can be determined based on the force distribution relationship under different trial directions, and then these multiple projected positions can be combined to form the candidate center of gravity position in the workpiece's reference coordinate system.

[0085] Multiple candidate masses and candidate centroid positions can be saved separately according to the corresponding effective time periods of the rigid body response. Then, the differences between candidate masses and the spatial distances between candidate centroid positions are statistically analyzed, grouping closely related and consecutively occurring candidate results into the same concentrated distribution range. This concentrated distribution range can be determined by the maximum permissible deviation range of the candidate results, or by the median value and dispersion of the candidate results. Candidate masses or candidate centroid positions that significantly deviate from other candidate results can be considered anomalous results, and their weight in the final calculation is reduced.

[0086] When multiple candidate masses are concentrated within the same mass range, the average, median, or weighted average of these candidate masses within that range can be used to determine the mass of the long-length flexible workpiece. The weight of each candidate mass can be determined based on the consistency between force changes and pose changes within the corresponding effective time period; the higher the consistency, the greater the weight of the corresponding candidate mass. Similarly, the initial center of gravity position can be determined based on multiple candidate center of gravity positions within the concentrated distribution range, with priority given to candidate center of gravity positions generated during effective time periods with higher consistency in rigid body response and smaller relative displacements at both ends.

[0087] Because long, flexible workpieces may exhibit residual bending even during in-phase impedance testing, this bending deformation can alter the instantaneous force distribution ratio between the two robot ends, causing the calculated candidate center of gravity positions to shift spuriously along the workpiece's bending direction. Therefore, after generating candidate center of gravity positions, they are arranged according to the time sequence of the effective response periods of each rigid body, and the direction and distance of change of adjacent candidate center of gravity positions are compared to determine the migration direction of the candidate center of gravity positions.

[0088] The migration direction of a candidate center of gravity position can be represented as the direction from the previous candidate center of gravity position to the next candidate center of gravity position. If the candidate center of gravity position moves towards the first robot side within multiple consecutive valid time periods, its migration direction is determined to point towards the first gripping position. If the candidate center of gravity position moves continuously towards the second robot side, its migration direction is determined to point towards the second gripping position. If the candidate center of gravity position fluctuates irregularly within a small range, it can be considered as normal computational fluctuation and not directly judged as bending deformation interference.

[0089] Matching the migration direction of the candidate center of gravity position with the relative displacement direction and response hysteresis relationship corresponding to the bending response characteristics refers to determining whether the change in the candidate center of gravity position corresponds to the workpiece bending deformation in both direction and time. First, the main bending direction of the workpiece is determined based on the relative displacement of the two robot ends during the anti-phase impedance test. Second, the time required for the relative displacement to reach a significant change after the bending action is determined based on the response hysteresis relationship. This time relationship is then mapped to the generation period of the candidate center of gravity position to determine whether the candidate center of gravity position migrates in the same direction after the bending response occurs.

[0090] When the migration direction of the candidate center of gravity is consistent with the relative displacement direction, and the migration time of the candidate center of gravity matches the response hysteresis relationship, it can be considered that the change in the candidate center of gravity is mainly caused by flexible deformation, rather than a change in the actual center of gravity of the workpiece. In this case, the corresponding candidate center of gravity is marked as a deformation interference position. For example, after the workpiece bends towards the first robot, the force on the end effector of the first robot temporarily increases, causing the calculated candidate center of gravity to move synchronously towards the first robot. In this case, the candidate center of gravity can be identified as the result of bending deformation interference.

[0091] When performing reverse compensation for deformation interference locations, the compensation direction and degree can be determined based on the relative displacement change over the corresponding time period. The compensation direction is opposite to the migration direction of the candidate center of gravity due to bending. The larger the relative displacement change, the more significant the impact of bending deformation on the force distribution relationship, and a larger compensation amount can be used. When the relative displacement change is small, a smaller compensation amount can be used. The compensation ratio can be predetermined through calibration tests on the same type of workpiece, or it can be determined online based on the correspondence between the migration of the candidate center of gravity and the relative displacement change over multiple trial periods.

[0092] For example, when a candidate center of gravity shifts towards the first gripping position due to the workpiece bending towards the first robot, the candidate center of gravity position is moved in the opposite direction towards the second robot based on the relative displacement change, thereby offsetting the false shift caused by the bending deformation. After compensation, the compensated candidate center of gravity position is compared again with other undisturbed candidate center of gravity positions. When multiple compensated candidate center of gravity positions re-enter the same concentrated distribution range, the final center of gravity position can be determined based on this concentrated distribution range.

[0093] If a candidate centroid position still deviates significantly from other candidate positions after compensation, it indicates that clamping slippage, sensor malfunction, or cooperative control error may exist simultaneously during that period, and the candidate centroid position can be excluded. The final centroid position can be determined jointly by the candidate centroid positions not marked as deformation interference positions and the candidate centroid positions that enter the concentrated distribution range after compensation.

[0094] Determining the flexibility state of a long flexible workpiece based on its bending response characteristics involves utilizing the loading, holding, and recovery relationships between the force difference and relative displacement during the reverse impedance test to determine the degree of deformation, recovery capability, and flexibility response bias of the workpiece when subjected to opposing forces at both ends. The flexibility state is not determined solely by the maximum value of the relative displacement, but rather by a comprehensive assessment combining the path of relative displacement change during the increase and decrease of the force difference, the remaining deformation after the test is lifted, and the direction and time relationship of the responses at both ends.

[0095] Specifically, the sequence of force difference changes in the bending response characteristics can be obtained chronologically. The period when the force difference begins to increase continuously from the baseline value is defined as the bending loading stage. During this stage, the opposing forces applied by the first and second robots gradually intensify, and the workpiece gradually undergoes bending deformation between the two gripping positions. The period when the force difference reaches its peak and then begins to decrease continuously until it recovers to the baseline range before the test is defined as the bending recovery stage. During this stage, the change in antiphase impedance is gradually removed or reduced, the elastic force accumulated inside the workpiece begins to be released, and the relative displacement gradually recovers.

[0096] To avoid errors in stage division caused by local fluctuations in the stress difference sequence, the stress difference can be required to maintain the same direction of change across multiple consecutive sampling times. When the stress difference briefly reverses direction but does not exceed the allowable fluctuation range, the corresponding data is still retained in the current stage. When the stress difference reaches its peak and remains stable for a certain period of time, this time range can be regarded as the holding period between the loading stage and the recovery stage, and the relative displacement at the beginning and end of the holding period is recorded respectively.

[0097] After determining the bending loading stage and the bending recovery stage, multiple identical or similar force difference values ​​are selected in both stages, and the relative displacements corresponding to these force difference values ​​are obtained in the loading and recovery stages, respectively. For example, when the force difference gradually increases, the first relative displacement corresponding to a certain force difference is recorded; when the force difference gradually decreases and the same force difference is reached again, the corresponding second relative displacement is recorded. Comparing the second relative displacement with the first relative displacement can determine whether the deformation path of the workpiece is consistent under the same external action level.

[0098] When the relative displacements corresponding to the bending loading stage and the bending recovery stage are basically the same under the same force difference, it indicates that the deformation and recovery process of the workpiece is relatively symmetrical, the bending energy can be released in time, and the degree of bending hysteresis is low. When the relative displacement corresponding to the recovery stage is significantly greater than that corresponding to the loading stage, it indicates that the force difference has decreased, but the workpiece still maintains a large bending deformation, and there is a significant delay in deformation recovery. The larger the difference in relative displacement between the two stages, the higher the degree of bending hysteresis.

[0099] The degree of bending hysteresis can be determined jointly based on the relative displacement differences under multiple stress levels. In practice, multiple comparison points can be selected below the peak stress difference, and the relative displacement differences at each comparison point during the loading and recovery phases can be calculated. The degree of bending hysteresis can then be determined based on the average, maximum, or cumulative value of these differences. Alternatively, the interval between the loading and recovery paths formed by the stress differences and relative displacements can be used as the degree of bending hysteresis. Using multiple comparison points can reduce the impact of noise and instantaneous vibration at a single sampling point on the judgment results.

[0100] After the bending recovery phase, the relative displacement between the ends of the first and second robots continues to be monitored. When the force difference has recovered to the pre-test reference range, and a relative displacement still exists between the two robot ends, this relative displacement is defined as the remaining relative displacement. The reference range can be determined based on the average value of the force difference and the allowable fluctuation before the start of the reverse impedance test. When the force difference enters the reference range and remains there for a preset time, the reverse external force is considered to have been substantially relieved.

[0101] The residual relative displacement characterizes the amount of deformation that the workpiece failed to recover immediately after the reverse impedance test. A small residual relative displacement that continues to decay to near zero indicates that the workpiece's bending deformation is primarily recoverable elastic deformation. A large residual relative displacement, or one that persists for an extended period during subsequent handling, indicates significant slow recovery, localized flexibility accumulation, or continuous deformation. To eliminate the influence of robot control errors, it is possible to simultaneously check whether the target poses of the two robot ends have been restored to their pre-test coordination. Only when the robot control commands have been restored and the actual relative displacement still exists should this relative displacement be considered the workpiece's residual relative displacement.

[0102] When determining the dominant side of the flexible response based on the response hysteresis relationship, the directional information of force difference and relative displacement can be retained, and the force changes, displacement changes, and peak occurrence times of the first and second robot sides can be analyzed separately. If the displacement change of the first robot side continues to increase after the force change, and its displacement peak has a long hysteresis time relative to the peak of the force difference, while the remaining relative displacement direction points towards the first robot side, then the first robot side can be determined as the dominant side of the flexible response. If the above response mainly occurs on the second robot side, then the second robot side can be determined as the dominant side of the flexible response.

[0103] The dominant side of the flexible response is primarily used to indicate that the flexible response of a long, flexible workpiece is not uniformly distributed between two gripping positions, but rather more prone to bending deformation to one side. Causes of lateral flexible response can include non-uniform workpiece cross-sections, local thickness variations, differences in internal structure, localized flexibility near the gripping position, and different clamping states at both ends. By combining the response direction and response lag time, it is possible to avoid misidentifying robot synchronization errors as lateral flexibility based solely on the displacement magnitude at a single moment.

[0104] After determining the degree of bending hysteresis, residual relative displacement, and dominant side of the flexible response, the flexible state of long-length flexible workpieces can be classified. When the change paths of the bending loading stage and the bending recovery stage are basically the same, the degree of bending hysteresis is low, and the residual relative displacement is small after the force difference is recovered, the workpiece is identified as being in an elastic bending state. Workpieces in this state can recover relatively quickly after the reverse action is removed, and their flexible response is usually relatively balanced between the two gripping positions.

[0105] When the bending hysteresis reaches the preset eccentricity judgment range, and the response hysteresis and residual relative displacement are mainly concentrated on the first robot side or the second robot side, the workpiece is determined to be in an eccentric flexible state. An eccentric flexible state indicates that one side of the workpiece is more prone to deformation or has a slower recovery speed. For this state, subsequent parameter configurations can reduce the load share borne by the side with dominant flexible response, or appropriately reduce the virtual stiffness of that side and increase virtual damping to reduce local stress concentration.

[0106] When the bending hysteresis is high, and there is still a residual relative displacement exceeding the allowable value after the force difference recovers to the reference range, or when the residual relative displacement does not decrease significantly in subsequent monitoring time, the workpiece is identified as being in a state of continuous deformation. This state indicates that the bending response of the workpiece cannot be eliminated within the normal recovery time, and may involve strong flexibility, localized plastic deformation, clamping slippage, or structural anomalies. In this state, the motion speed and acceleration of both robots can be reduced, and the allowable relative displacement and cooperative internal forces can be limited.

[0107] In practical implementation, the judgment ranges for bending hysteresis and residual relative displacement can be pre-set for long flexible workpieces of different sizes and materials. The judgment ranges can also be updated based on the response results of the same workpiece in multiple reverse-phase impedance tests. When the flexibility states obtained from multiple tests are inconsistent, the state with higher bending hysteresis or larger residual relative displacement can be prioritized, thereby improving the safety of subsequent collaborative handling control.

[0108] By using the above method, the actual flexible state of the workpiece can be identified from the loading path, recovery path, residual state and lateral response of the bending deformation, avoiding simple classification based solely on the maximum displacement or maximum force, and providing a more accurate basis for the configuration of load sharing, virtual stiffness, virtual damping and motion limit parameters between the two robots.

[0109] Step 106 specifically involves configuring the load sharing ratio, virtual stiffness parameters, virtual damping parameters, and motion restriction parameters of the first robot and the second robot based on the mass, center of gravity position, and flexibility state.

[0110] The load-sharing ratio, virtual stiffness parameters, and virtual damping parameters of the first and second robots are configured according to their mass, center of gravity position, and flexibility. This means first determining the basic load that the two robots should bear based on the static load distribution of the workpiece, and then controlling the transfer of the basic load according to the degree of lateral deviation and recovery ability of the workpiece's bending deformation. During the load transfer process, the impedance control parameters of the two robots are adjusted synchronously. In this way, while the two robots jointly support the total mass of the workpiece, the local load borne by the side with the dominant flexible response is reduced, avoiding the exacerbation of workpiece bending caused by simply distributing the load according to the center of gravity position.

[0111] Specifically, the required support force from both robots is determined based on the identified workpiece mass, and the basic load-sharing ratio is determined based on the distance between the workpiece's center of gravity and the first and second gripping positions. The line connecting the two gripping positions can be used as the load distribution direction, and the required support force at both ends is determined based on the position of the center of gravity relative to the two gripping positions. The closer the center of gravity is to the first robot, the greater the basic load-sharing ratio for the first robot. The closer the center of gravity is to the second robot, the greater the basic load-sharing ratio for the second robot. The sum of the basic load-sharing ratios of the two robots corresponds to the total support load of the workpiece.

[0112] The basic load-sharing ratio primarily reflects the static load distribution caused by the workpiece's mass and center of gravity, without directly considering the additional load caused by flexible bending. When the workpiece is in an elastic bending state and the flexible response on both sides is relatively balanced, the basic load-sharing ratio can be directly used, or only minor dynamic adjustments can be made. When the workpiece is in a lateral flexible state or a state of continuous deformation, the basic load-sharing ratio needs to be further adjusted based on the dominant side of the flexible response, the degree of bending hysteresis, and the remaining relative displacement.

[0113] The dominant flexible response side is used to determine the robot to be unloaded. If one side of the first robot is determined to be the dominant flexible response side, it indicates that this side has significant displacement lag, slow recovery, or residual relative displacement during the reverse impedance test; therefore, the first robot is determined to be the robot to be unloaded. If one side of the second robot is the dominant flexible response side, then the second robot is determined to be the robot to be unloaded. Unloading here does not mean removing the robot's support from the workpiece, but rather transferring part of the load borne by the robot to another robot while maintaining a common clamping state.

[0114] The allowable load transfer amount limits the maximum load share transferred from the robot to be unloaded to another robot. It can be determined based on the current basic load-sharing ratio of the robot to be unloaded, the remaining load-bearing capacity of the other robot, the degree of bending hysteresis, and the remaining relative displacement. When the degree of bending hysteresis is low and the remaining relative displacement is small, it indicates that the workpiece has good recovery ability, and a smaller allowable load transfer amount can be set. When the degree of bending hysteresis is high or the remaining relative displacement is large, it indicates that the dominant side of the flexible response is prone to continuous deformation accumulation, and the allowable load transfer amount can be appropriately increased within the load-bearing capacity of the other robot.

[0115] When determining the allowable load transfer amount, it is also necessary to check the end effector allowable load, joint drive torque margin, load-bearing capacity in the current posture, and clamping stability of the other robot. The allowable load transfer amount must not cause the other robot to exceed the corresponding load limit, nor should it cause the support force of the robot to be unloaded to be lower than the minimum support force required to maintain clamping stability. When the transfer amount calculated according to the flexibility state exceeds the load limit, the transferable load corresponding to the load limit is determined as the actual allowable load transfer amount.

[0116] During load transfer, the combined support force of the first and second robots is kept within a preset range. This combined support force can be obtained by combining the end forces of the two robots along the main support direction of the workpiece. Its preset range is determined based on the workpiece mass, current acceleration, and allowable support force fluctuations. By limiting the combined support force, it can be ensured that the load reduction of one robot roughly corresponds to the load increase of the other robot, avoiding insufficient overall support for the workpiece or excessive support force applied by both robots simultaneously during load transfer.

[0117] Virtual stiffness characterizes the strength of the restoring effect of a robot's end effector when it deviates from the target pose. Reducing the virtual stiffness of the robot to be unloaded weakens the constraint of its end effector on local workpiece pose changes, thus reducing the additional force it bears. Increasing the virtual stiffness of the other robot allows it to bear more support load while maintaining the target pose. Therefore, by adjusting the virtual stiffness of the two robots in opposite directions, the workpiece load can be gradually transferred from the robot to be unloaded to the other robot.

[0118] Virtual stiffness can be adjusted in a step-by-step manner. For example, the allowable load transfer can be divided into multiple transfer stages. After each stage is completed, the actual forces and relative displacements of the two robot end effectors are reacquired. The next transfer stage is only initiated if no excessive combined support force, increased relative displacement, or abnormal clamping state occurs in the current stage. Compared to modifying the load-sharing ratio all at once, gradually adjusting the virtual stiffness can reduce sudden workpiece swaying, load impacts, and control conflicts between the two robots.

[0119] In practice, the virtual stiffness of the robot to be unloaded can be reduced by a preset step size, and the virtual stiffness of the other robot can be increased by a corresponding step size. After each adjustment, the actual load share borne by the two robots' ends is obtained. When the actual load transfer amount has not yet reached the target transfer amount for the current stage, complementary stiffness adjustments continue. When the actual load transfer amount has reached the target value, stiffness adjustments are paused and the system enters a response stability assessment phase.

[0120] Virtual damping is used to suppress oscillations during robot end-effector pose changes and load transfer. When a load is transferred from one robot to another, the force difference and relative displacement between the two robot end-effectors may fluctuate. In this case, the virtual damping parameters of the two robots are adjusted according to the actual changes in force difference and relative displacement to limit the load transfer speed and promote the attenuation of workpiece bending response.

[0121] When the force difference increases rapidly after load transfer, or when the relative displacement changes repeatedly, the virtual damping of the corresponding robot can be increased to reduce the adjustment speed of the robot's end effector and prevent the load from being transferred back and forth between the two robots. When the force difference changes relatively smoothly, but the recovery speed of the remaining relative displacement is slow, the virtual damping of the side with dominant flexible response can be appropriately increased to absorb the bending vibration on that side. When excessive virtual damping causes the load transfer to stagnate, the virtual damping of the other robot can be appropriately reduced to enable it to continue to bear the target transfer load.

[0122] The allowable range of stress difference can be determined based on the theoretical stress difference corresponding to the basic load sharing ratio and the allowable deviation. The allowable range of relative displacement can be determined based on workpiece size, material flexibility, clamping accuracy, and handling safety requirements. After each stiffness and damping adjustment, the actual stress difference and relative displacement are compared with the corresponding allowable ranges. Only when the stress difference no longer continues to increase, the relative displacement gradually decreases, and both enter the corresponding allowable ranges, is the current load migration state considered to meet the requirements.

[0123] Once the force difference and relative displacement return to their respective allowable ranges, the actual loads currently borne by the first and second robots are obtained, and the proportion of the current load of the two robots to the combined support force is calculated. This proportion is then determined as the target load-sharing ratio. Simultaneously, the virtual stiffness parameters and virtual damping parameters at the corresponding moment are saved as the target impedance configuration parameters for the current workpiece, to be used in subsequent collaborative handling sections.

[0124] For example, when the workpiece's center of gravity is biased towards the first robot, the static torque balance relationship might require the first robot to bear a larger base load. However, if the reverse impedance test indicates that the first robot side is also the side dominated by flexible response, a high load ratio cannot be maintained directly in the long term. In this case, provided the second robot still has a load-bearing margin, the virtual stiffness of the first robot can be reduced and the virtual stiffness of the second robot increased, transferring part of the load borne by the first robot to the second robot. The amount of transfer should be limited to reducing the relative displacement on the first robot side without causing overload on the second robot.

[0125] If the relative displacement does not decrease during load migration, or if the dominant flexible response shifts from the first robot to the second robot, it indicates that the migration amount is too large or the stiffness configuration is unreasonable. In this case, migration can be stopped, the stiffness and damping parameters of the previous stable phase restored, and the load sharing ratio corresponding to the previous stable phase determined as the target load sharing ratio. By maintaining the stable configuration, excessive unloading can prevent the workpiece from bending to the other side.

[0126] For elastic bending, the target load sharing ratio can be made close to the base load sharing ratio, and virtual stiffness and moderate virtual damping can be used to quickly restore the posture. For lateral flexibility, the target load ratio and virtual stiffness on the side with dominant flexible response can be appropriately reduced, and the virtual damping on that side can be increased. For continuous deformation, the allowable load transfer speed can be limited, while the overall virtual stiffness variation of the two robots can be reduced to avoid the workpiece being subjected to large additional internal forces.

[0127] By employing the above configuration, while meeting the basic load-bearing requirements for workpiece mass and center of gravity formation, a flexible state can be introduced into the load distribution process between the two robots. Virtual stiffness controls the load migration direction, and virtual damping controls the load migration speed and response oscillations. This reduces the force concentration on the dominant side of the flexible response, decreases workpiece bending and the cooperative internal forces between the two robots, and improves the stability and safety of handling long, flexible workpieces.

[0128] To further illustrate the technical effects of this application, a specific embodiment and its test results are presented below.

[0129] Two six-degree-of-freedom industrial robots were used to grip a flexible aluminum alloy sheet with a length of 3.2 meters and a width of 0.45 meters. The sheet had an actual mass of 36.0 kg, and its center of gravity was located at 52.5% of the line connecting the first and second gripping positions. Six-dimensional force sensors were installed at the end effectors of both robots, with a data sampling frequency of 500 Hz. The system designated the horizontal uniform-speed transport section as the trial motion section. First, the two robots were controlled to synchronously generate a 0.8 mm in-phase pose change along the transport direction. Then, they were controlled to generate opposite pose changes of 0.6 mm along the transverse direction of the sheet. The effective time period of the rigid body response was determined based on the force components, pose synchronization components, and response phase relationship during the in-phase impedance test, and candidate masses and candidate center of gravity positions were calculated. Subsequently, bending response characteristics were generated based on the force difference, relative displacement, and peak time difference during the out-of-phase impedance test. The sheet was identified as being in a side-biased flexible state by the first robot, and the candidate center of gravity position affected by bending was compensated in reverse.

[0130] Thirty consecutive handling tests were conducted using the above method. The average identified mass was 35.7 kg, the average relative error of the mass was 0.83%, the average identified center of gravity position was 52.1%, and the position deviation was 0.4% of the gripping distance. Without deformation interference compensation, the maximum deviation of the center of gravity was 4.6%, which was reduced to 1.2% after compensation. Based on the identification results, the load sharing ratio of the two robots was adjusted from the initial 50:50 to 47:53. After simultaneously adjusting the virtual stiffness and virtual damping parameters, the peak force on the first robot side decreased from 238 N to 196 N, and the maximum relative displacement of the two robots' ends decreased from 7.4 mm to 3.1 mm. No gripping slippage, continuous oscillation, or workpiece posture exceeding limits occurred during the handling process, indicating that the proposed solution can improve the load identification accuracy of flexible workpieces and the stability of dual-robot collaborative handling.

[0131] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for intelligent identification and adaptive configuration of robot end-effector loads, characterized in that, include: Obtain the collaborative transport trajectory when the first robot and the second robot jointly clamp a long flexible workpiece; The trial motion segment is determined from the cooperative transport trajectory, and the first robot and the second robot are controlled to perform in-phase impedance testing and out-of-phase impedance testing respectively in the trial motion segment; wherein, the in-phase impedance testing refers to applying virtual impedance changes in the same direction and time to the first robot and the second robot, so that the end effectors of the two robots produce pose changes in the same direction; the out-of-phase impedance testing refers to applying virtual impedance changes in opposite directions to the first robot and the second robot, so that the end effectors of the two robots produce relative displacement; The force and pose changes of the two robot ends during the in-phase impedance test are obtained, and the rigid body motion response characteristics of the long flexible workpiece are generated based on the consistent relationship between the force and pose changes. The force difference, relative displacement and response hysteresis relationship of the two robot ends during the anti-phase impedance test are obtained, and the bending response characteristics of the long flexible workpiece are generated. The mass and center of gravity of the long flexible workpiece are determined based on the rigid body motion response characteristics, and the flexibility state of the long flexible workpiece is determined based on the bending response characteristics. Based on the mass, center of gravity position, and flexibility state, the load sharing ratio, virtual stiffness parameters, virtual damping parameters, and motion restriction parameters of the first robot and the second robot are configured respectively.

2. The method according to claim 1, characterized in that, Determining the trial movement segment from the cooperative transport trajectory includes: The cooperative transport trajectory is divided into multiple candidate motion segments; Based on the changes in motion acceleration corresponding to each candidate motion segment, the force margin at the end of the two robots, and the workpiece posture deviation margin, the test motion segment that meets the preset test conditions is determined.

3. The method according to claim 1, characterized in that, The in-phase impedance test includes: A virtual impedance change in the same direction and time is applied to the first robot and the second robot, causing the end effectors of the two robots to produce a small pose change in the same direction, while keeping the predetermined transport direction of the long flexible workpiece unchanged.

4. The method according to claim 1, characterized in that, The reverse impedance test includes: Virtual impedance changes in opposite directions are applied to the first and second robots to cause relative displacement at the ends of the two robots. The motion amplitude of the ends of the two robots is adjusted according to the relative displacement to limit the overall positional offset of the long flexible workpiece.

5. The method according to claim 1, characterized in that, The process of generating rigid body motion response characteristics of the long-sized flexible workpiece based on the consistent relationship between force changes and pose changes includes: Extract the co-directional component of the force change at the end effector of the two robots and the synchronous component of the pose change; The rigid body motion response characteristics are generated based on the same-direction component, the synchronous component, and the response phase relationship between the two.

6. The method according to claim 1, characterized in that, The generation of the bending response characteristics of the long-sized flexible workpiece includes: The force difference is determined based on the force changes at the ends of the first and second robots, and the relative displacement is determined based on the pose changes at the ends of the two robots. Determine the response hysteresis relationship between the time when the force difference reaches its peak value and the time when the relative displacement reaches its peak value; The bending response characteristics are generated based on the variation range of the force difference, the variation range of the relative displacement, and the response hysteresis relationship.

7. The method according to claim 1, characterized in that, Determining the mass and center of gravity position of the long flexible workpiece based on the rigid body motion response characteristics includes: Based on the degree of consistency between the force changes and pose changes of the two robot ends in the rigid body motion response characteristics, multiple effective time periods of rigid body response are determined. Multiple candidate masses are generated based on the resultant force and common acceleration of the two robot ends during the effective time period of each rigid body response. Based on the candidate mass, the force distribution relationship between the two robot ends, and the grasping position, the corresponding candidate center of gravity position is generated; A concentrated distribution range of multiple candidate masses and candidate center of gravity positions is determined, and the mass and center of gravity position of the long flexible workpiece are determined based on the concentrated distribution range.

8. The method according to claim 7, characterized in that, After generating the corresponding candidate centroid positions, the process also includes: The candidate centroid positions are arranged in chronological order according to the effective time period of each rigid body response, and the migration direction of the candidate centroid positions is determined. Match the migration direction with the relative displacement direction and response hysteresis relationship corresponding to the bending response characteristics; Candidate centroid locations whose migration direction changes synchronously with the bending response are marked as deformation disturbance locations; The deformation interference position is compensated in reverse according to the relative displacement change in the corresponding time period, and the center of gravity position of the long flexible workpiece is determined according to the candidate center of gravity position after compensation.

9. The method according to claim 1, characterized in that, Determining the flexibility state of the long-length flexible workpiece based on the bending response characteristics includes: Based on the increasing and decreasing processes of the force difference in the bending response characteristics, the bending loading stage and the bending recovery stage are determined respectively. By comparing the relative displacements corresponding to the bending loading stage and the bending recovery stage under the same force difference, the degree of bending hysteresis is determined. After the force difference recovers to the reference range, the remaining relative displacement is determined, and the dominant side of the flexible response is determined according to the response hysteresis relationship; Based on the degree of bending hysteresis, the remaining relative displacement, and the dominant side of the flexible response, the elastic bending state, the lateral flexible state, or the continuous deformation state are determined.

10. The method according to claim 9, characterized in that, The step of configuring the load-sharing ratio, virtual stiffness parameters, and virtual damping parameters of the first robot and the second robot according to the mass, center of gravity position, and flexibility state includes: The basic load-sharing ratio of the two robots is determined based on the location of the center of gravity. Based on the dominant side of the flexible response, the degree of bending hysteresis, and the remaining relative displacement, determine the robot to be unloaded and the allowable load migration amount; While keeping the combined support force of the two robots within a preset range, the virtual stiffness of the robot to be unloaded is reduced, and the virtual stiffness of the other robot is increased, so that the allowable load transfer amount is gradually transferred to the other robot. Based on the force differences and relative displacement changes during the load migration process, the virtual damping parameters of the two robots are adjusted respectively, and the load sharing ratio when the force differences and relative displacement are restored to the corresponding allowable range is determined as the target load sharing ratio.

Images