Robot grasping control system for special-shaped workpieces based on visual guidance and dynamic force matching

Abstract

This invention relates to the field of vision technology in the robotics industry, and discloses a robot gripping control system for irregularly shaped workpieces based on vision guidance and dynamic force matching. The system includes an image acquisition module, a model reconstruction module, a pose optimization module, and a collaborative control module. The system acquires visual point cloud data of the material to be gripped to reconstruct a parameterized 3D mesh model. It uses Boolean interference operations between the mesh model and a preset robotic claw envelope model to retrieve the optimal topological nested pose and generate a normal constraint threshold. The collaborative control module anchors the gripping point based on static geometric boundaries and drives the robotic claw to grip through a closed-loop counter-pressure mechanism of real-time contact pressure and the normal constraint threshold. This invention transforms the nonlinear dynamic response process into a topological matching process within geometric space, avoiding material displacement caused by lag in the mechanical transmission chain response and ensuring the structural integrity of the thin-walled reel under vertical placement conditions.

Description

Technical Field

[0001] This invention relates to the field of vision technology in the robotics industry, and in particular to a robot gripping control system for irregularly shaped workpieces based on vision guidance and dynamic force matching. Background Technology

[0002] In current semiconductor packaging and testing production, the mainstream solution is to use a vision-guided system to drive a robotic arm to pick up and place materials from reels. Existing technologies typically rely on visual image recognition of material characteristics to determine the tension parameters of the pick-up and place mechanism, thereby realizing the picking, placing, and transfer of materials. When dealing with vertically stored reels of materials with various specifications, the above solution faces the constraint of mechanical transmission response lag and nonlinear deformation of the material contact surface. Due to the physical time delay between signal processing and mechanical action response of the mechanical transmission system, the control closed loop exhibits a posteriori properties. When the sensor captures the material displacement characteristics, the static friction of the material contact surface has often been converted into dynamic friction, making it difficult for subsequent torque compensation to reconstruct static equilibrium in a timely manner.

[0003] Furthermore, for structurally fragile reel materials, simply increasing the clamping torque can easily cause stress concentration in the contact area, inducing local physical damage to the reel. The inherent physical lag in the hardware transmission layer is difficult to overcome, and the feedback method that relies heavily on software control strategies also has its shortcomings. Chinese invention patent application CN120828418A discloses a vision recognition-based robotic gripping control system. It extracts features through multimodal fusion and relies on dynamic feedback closed-loop to collect tactile perception data in real time to correct the gripping force. Analysis of the underlying mechanism reveals that it relies on tactile feedback after contact, supplemented by PID adjustment control logic. It has not broken away from the posterior framework of contact first, perception then compensation. When facing thin-walled reels that are prone to buckling and deformation in semiconductor scenarios, the physical time for the tactile sensor to capture pressure changes and transmit them to the control hub and then for the drive unit to output compensation torque is much longer than the critical value for the reel contact surface to go from elastic deformation to plastic damage or from static friction to dynamic friction. The control preset premise is fundamentally mismatched with the actual physical response limit, resulting in dynamic torque compensation being slower than the transient instability process of the material.

[0004] Therefore, how to overcome the response limit of feedback control and establish a grasping mechanism with geometric constraints before physical contact occurs is the technical problem to be solved by this invention. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a robot gripping and controlling system for irregularly shaped workpieces based on vision guidance and dynamic force matching, the system comprising: The image acquisition module acquires visual point cloud data of the material to be grabbed from the reel; The model reconstruction module reconstructs a parametric 3D mesh model of the reel material to be grasped based on visual point cloud data, and retrieves a preset 3D envelope model of the mechanical claw's kinematics. The pose optimization module, within the computer-aided design logic framework, uses Boolean interference intersection operations between a parametric 3D mesh model and a 3D envelope model of the mechanical gripper's kinematics to retrieve the geometrically constrained nested poses that maximize the topological overlap volume without exceeding the theoretical yield boundary of the material to be gripped. Based on the contact surface area corresponding to the topological overlap volume and the preset elastic modulus of the material to be gripped, it calculates and generates the normal constraint threshold and outputs the static geometric boundary coordinate system. The collaborative control module anchors the gripping point based on the static geometric boundary coordinate system and performs closed-loop offsetting between the real-time contact pressure at the end of the mechanical claw and the normal constraint threshold. By driving the mechanical claw to grip the reel material to be gripped, the dynamic response process of gripping is converted into a topological matching process in geometric space, which is used to limit the axial displacement of the reel material to be gripped at the moment of contact.

[0006] Preferably, when the pose optimization module searches for the optimal contact surface, it applies the criterion of minimizing the variance of the normal gradient of the interference surface; the collaborative control module adjusts the mechanical claw according to the optimal contact surface to make the clamping stress evenly distributed on the surface of the reel material to be gripped, so as to control the normal stress generated by clamping to be lower than the preset elastic limit of the reel material to be gripped, and maintain the structural stability of the reel material to be gripped under vertical placement conditions.

[0007] Preferably, the model reconstruction module extracts the material specification feature vector based on the center aperture characteristics and end face flatness data of the material to be grasped from the reel material, and matches the corresponding elastic modulus parameters and geometric structure parameters from the preset workpiece library to support the dynamic fitting of the parametric three-dimensional mesh model.

[0008] Preferably, when the pose optimization module retrieves the nested pose of geometric constraints, it calculates the minimum overlapping area required to establish stable clamping in the virtual geometric space using the material surface roughness coefficient and static friction factor, and corrects the static geometric boundary coordinate system based on the minimum overlapping area.

[0009] Preferably, the pose optimization module defines the topological interference volume by solving the intersection of the parametric 3D mesh model and the 3D envelope model of the mechanical claw's kinematics in 3D space. Topological interference volume The calculation method is as follows: ,in, For topological interference volume, The spatial region occupied by the parametric 3D mesh model. The spatial region occupied by the three-dimensional envelope model of the mechanical claw's kinematics.

[0010] Preferably, the collaborative control module includes a pressure feedback unit. After the mechanical gripper comes into contact with the material to be gripped from the reel, the pressure feedback unit collects the real-time contact pressure at the end of the mechanical gripper and performs a difference calculation between the real-time contact pressure and the normal constraint threshold. When the real-time contact pressure deviates from the normal constraint threshold, a tension force compensation signal is generated.

[0011] Preferably, the system also includes an offset monitoring module, which uses a visual feature tracking algorithm to monitor the axial displacement of the material to be gripped on the reel during the movement process; when the axial displacement exceeds a preset safety deviation threshold, the collaborative control module drives the mechanical claw to work together with the limit locking component, wherein the safety deviation threshold ranges from 0.5mm to 2.0mm.

[0012] Preferably, the collaborative control module also includes a motion compensation unit. The motion compensation unit calculates the dynamic inertial load based on the real-time acceleration of the mechanical gripper and the mass parameters of the material to be gripped from the reel, and performs feedforward correction on the driving torque of the mechanical gripper based on the dynamic inertial load in order to maintain the dynamic balance state during the gripping process.

[0013] Preferably, the image acquisition module includes a multi-angle industrial camera group and a supplementary lighting unit; before acquiring visual point cloud data, the image acquisition module filters out environmental reflection noise from the surface of the material to be grasped on the reel through a feature point matching algorithm, so as to improve the reconstruction accuracy of the parametric three-dimensional mesh model.

[0014] Preferably, the system converts the mechanical gripper into a parameterized constraint entity within a computer-aided design logic framework, and completes the geometric topological nesting of the parameterized three-dimensional mesh model before the physical gripping action is initiated, thereby achieving adaptive anti-offset control of multi-specification vertically placed reel materials in the semiconductor packaging and testing station.

[0015] The beneficial effects of this invention are: 1. In the dynamic force matching of the robot's grasping control of irregular workpieces, this invention transforms the traditional dynamic torque compensation process into a geometric topology addressing process in virtual space through parametric reconstruction of the three-dimensional mesh model and the actuator entity model. Since the system performs dynamic Boolean interference analysis in virtual space before physical contact occurs and locks the geometric interlock configuration that maximizes the topological interference volume, the end effector of the robot can directly anchor to the preset static geometric boundary. This approach avoids the inherent physical response lag of the mechanical transmission chain and prevents the phase transition from static friction to dynamic friction at the moment of grasping, thereby avoiding material slippage and the risk of material displacement caused by a posteriori torque compensation.

[0016] 2. By utilizing the optimization criterion of minimizing the normal gradient variance of the interference surface, this invention ensures the uniform distribution of clamping stress on the contact surface of the reel material. When handling thin-walled reel materials in semiconductor chip packaging and testing scenarios, this mechanism eliminates nonlinear stress concentration in local areas by precisely aligning the topology during the physical addressing stage. This linearization of stress distribution avoids local buckling deformation of the reel caused by stress peaks during the adaptive matching process of the clamping force, thus ensuring the structural integrity of the material throughout the entire process of vertical storage and handling.

[0017] 3. By coupling the kinematic envelope of the pick-and-place actuator with the real-time acquired 3D point cloud features, this invention constructs a parameterized constraint system that can adaptively accommodate materials of various specifications. The system no longer relies on preset fixed grasping poses, but dynamically generates a unique set of spatial contact coordinate points based on the topological boundary of each material. This interlocking mechanism based on geometric constraints enables a single actuator to stably handle vertically placed reels with different diameters, widths, and material hardness, reducing the extreme dependence on the physical feedback accuracy of the actuator and improving the operational reliability of the robot system in complex semiconductor material transfer scenarios. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a flowchart of the robot's grasping and control operation for irregularly shaped workpieces according to the present invention; Figure 2 This is a diagram showing the system modules and interaction of the vision and force matching system of the present invention. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, an embodiment or embodiment referred to herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention. An embodiment appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] This invention is described in detail with reference to the schematic diagrams. When describing the embodiments of this invention, for ease of explanation, the cross-sectional views of the device structure will be partially enlarged without adhering to the general scale. Moreover, the schematic diagrams are only examples and should not limit the scope of protection of this invention. In addition, in actual manufacturing, the three-dimensional spatial dimensions of length, width and depth should be included.

[0023] Furthermore, in the description of this invention, it should be noted that the terms such as "upper," "lower," "inner," and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or component referred to has a specific orientation, or is constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0024] Unless otherwise explicitly specified and limited, the terms installation, connection, and linking in this invention should be interpreted broadly. For example, they can refer to fixed connection, detachable connection, or integrated connection; similarly, they can refer to mechanical connection, electrical connection, or direct connection, or indirect connection through an intermediate medium, or internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0025] A vision-guided and dynamic force-matching-based robot gripping control system for irregularly shaped workpieces. The system includes: The image acquisition module acquires visual point cloud data of the material to be grabbed from the reel; The model reconstruction module reconstructs a parametric 3D mesh model of the reel material to be grasped based on visual point cloud data, and retrieves a preset 3D envelope model of the mechanical claw's kinematics. The pose optimization module, within the computer-aided design logic framework, uses Boolean interference intersection operations between a parametric 3D mesh model and a 3D envelope model of the mechanical gripper's kinematics to retrieve the geometrically constrained nested poses that maximize the topological overlap volume without exceeding the theoretical yield boundary of the material to be gripped. Based on the contact surface area corresponding to the topological overlap volume and the preset elastic modulus of the material to be gripped, it calculates and generates the normal constraint threshold and outputs the static geometric boundary coordinate system. The collaborative control module anchors the gripping point based on the static geometric boundary coordinate system and performs closed-loop offsetting between the real-time contact pressure at the end of the mechanical claw and the normal constraint threshold. By driving the mechanical claw to grip the reel material to be gripped, the dynamic response process of gripping is converted into a topological matching process in geometric space, which is used to limit the axial displacement of the reel material to be gripped at the moment of contact.

[0026] Preferably, when the pose optimization module searches for the optimal contact surface, it applies the criterion of minimizing the variance of the normal gradient of the interference surface; the collaborative control module adjusts the mechanical claw according to the optimal contact surface to make the clamping stress evenly distributed on the surface of the reel material to be gripped, so as to control the normal stress generated by clamping to be lower than the preset elastic limit of the reel material to be gripped, and maintain the structural stability of the reel material to be gripped under vertical placement conditions.

[0027] Preferably, the model reconstruction module extracts the material specification feature vector based on the center aperture characteristics and end face flatness data of the material to be grasped from the reel material, and matches the corresponding elastic modulus parameters and geometric structure parameters from the preset workpiece library to support the dynamic fitting of the parametric three-dimensional mesh model.

[0028] Preferably, when the pose optimization module retrieves the nested pose of geometric constraints, it calculates the minimum overlapping area required to establish stable clamping in the virtual geometric space using the material surface roughness coefficient and static friction factor, and corrects the static geometric boundary coordinate system based on the minimum overlapping area.

[0029] Preferably, the pose optimization module defines the topological interference volume by solving the intersection of the parametric 3D mesh model and the 3D envelope model of the mechanical claw's kinematics in 3D space. Topological interference volume The calculation method is as follows: ,in, For topological interference volume, The spatial region occupied by the parametric 3D mesh model. The spatial region occupied by the three-dimensional envelope model of the mechanical claw's kinematics.

[0030] Preferably, the collaborative control module includes a pressure feedback unit. After the mechanical gripper comes into contact with the material to be gripped from the reel, the pressure feedback unit collects the real-time contact pressure at the end of the mechanical gripper and performs a difference calculation between the real-time contact pressure and the normal constraint threshold. When the real-time contact pressure deviates from the normal constraint threshold, a tension force compensation signal is generated.

[0031] Preferably, the system also includes an offset monitoring module, which uses a visual feature tracking algorithm to monitor the axial displacement of the material to be gripped on the reel during the movement process; when the axial displacement exceeds a preset safety deviation threshold, the collaborative control module drives the mechanical claw to work together with the limit locking component, wherein the safety deviation threshold ranges from 0.5mm to 2.0mm.

[0032] Preferably, the collaborative control module also includes a motion compensation unit. The motion compensation unit calculates the dynamic inertial load based on the real-time acceleration of the mechanical gripper and the mass parameters of the material to be gripped from the reel, and performs feedforward correction on the driving torque of the mechanical gripper based on the dynamic inertial load in order to maintain the dynamic balance state during the gripping process.

[0033] Preferably, the image acquisition module includes a multi-angle industrial camera group and a supplementary lighting unit; before acquiring visual point cloud data, the image acquisition module filters out environmental reflection noise from the surface of the material to be grasped on the reel through a feature point matching algorithm, so as to improve the reconstruction accuracy of the parametric three-dimensional mesh model.

[0034] Preferably, the system converts the mechanical gripper into a parameterized constraint entity within a computer-aided design logic framework, and completes the geometric topological nesting of the parameterized three-dimensional mesh model before the physical gripping action is initiated, thereby achieving adaptive anti-offset control of multi-specification vertically placed reel materials in the semiconductor packaging and testing station.

[0035] Example 1: In the automated transfer of vertically placed reel materials in a semiconductor chip packaging and testing production line, the structural stability of multi-specification thin-walled reel materials in vertical storage is limited. Traditional clamping systems based on physical offset a posteriori feedback, when visually detecting material slippage, have already converted static friction at the contact surface into dynamic friction. At this point, the tension compensation signal applied by the system's pressure feedback loop often lags behind the physical response time of the mechanical transmission chain. This feedback lag and the inherent physical contradiction between the nonlinear force deformation of the reel contact surface cause nonlinear stress concentration in localized areas, leading to edge curling and physical drop damage to the semiconductor reel. The image acquisition module acquires visual point cloud data of the reel material to be gripped. The model reconstruction module reconstructs a parametric 3D mesh model of the reel material based on the visual point cloud data, retrieves a preset 3D kinematic envelope model of the mechanical gripper, and the pose optimization module, within the computer-aided design logic framework, optimizes the spatial area occupied by the 3D kinematic envelope model of the mechanical gripper. For reference, the spatial area occupied by the relative parametric 3D mesh model To solve for the intersection in three-dimensional space, the system calculates the interference overlap region in real time based on the following formula: ,in, For topological interference volume, The spatial region occupied by the parametric 3D mesh model. The spatial region occupied by the three-dimensional envelope model of the mechanical claw's kinematics.

[0036] The system applies the criterion of minimizing the variance of the normal gradient of the interferometric surface. Based on the theory of interface morphology in differential geometry, the spatial dispersion of the contact surface normal vector directly reflects the nonlinear stress concentration tendency under the clamping state of the entity. The pose optimization module extracts a preset number of discrete mesh vertices within the estimated topological coincident plane, calculates the spatial angle between the surface normal vector at each discrete mesh vertex and the main clamping axis of the manipulator, and constructs an objective function guided by minimizing the statistical variance of the spatial angle using the six-degree-of-freedom pose matrix of the manipulator's kinematic three-dimensional envelope model as the iterative parameter. The gradient descent algorithm is used to update the six-degree-of-freedom pose matrix one by one. In the initialization stage of the underlying operation, the basic translation iteration step size is set to 0.1 mm and the rotation iteration step size is set to 0.5 degrees. In each iteration, the arithmetic variance of the angle values ​​of all discrete points is calculated. If the variance value of the current iteration cycle is less than that of the previous cycle, the iteration step size is multiplied by an acceleration factor of 1.2; if it is greater than that of the previous cycle, the iteration step size is multiplied by a decay factor of 0.5 and the iteration is reversed. The iteration terminates when the difference in variance between two adjacent iterations is lower than the convergence tolerance. The spatial matrix alignment system performs a directional search in the virtual geometric space to maximize the topologically overlapping volume without exceeding the geometric constraints of the theoretical yield boundary of the material to be grasped. The collaborative control module calculates the normal constraint threshold based on the contact surface area corresponding to the topologically overlapping volume and the preset elastic modulus of the material to be grasped. Based on the linear elastic contact deformation mechanism, the equivalent normal reaction force of the solid interface is linearly positively correlated with the spatial topological indentation depth. The pose optimization module extracts 20 discrete sampling points uniformly distributed within the topologically overlapping region from the parameterized 3D mesh model. The spatial Euclidean distance between the inner and outer surfaces of the mesh is measured along the surface normal direction of each sampling point. After removing the maximum and minimum values ​​using a sorting algorithm, the arithmetic mean of the remaining 18 distance values ​​is calculated as the characteristic wall thickness of the overlapping region. Combined with the equivalent depth calculation of the interference volume, the specific formula for calculating the normal constraint threshold is as follows: ,in, λ is the normal constraint threshold; λ is the dimensionless geometric stiffness correction coefficient, ranging from 0.85 to 1.15; E is the preset elastic modulus. For topological interference volume; Characteristic wall thickness.

[0037] To ensure the feasibility of the project, the dimensionless geometric stiffness correction coefficient in the formula is not arbitrarily chosen, but is generated offline by the physical probe detection program at the initial power-on stage of the system: the mechanical gripper is slowly pressed into a standard rigid force measuring block under no-load conditions, and the actual physical rebound force value when the displacement of one side of the gripper finger reaches 1mm is recorded. This value is divided by the theoretical linear elasticity calculation result, and a one-dimensional lookup table with a pre-fixed image index step size of 0.05 is generated. During normal operation, the microprocessor directly addresses and calls the coefficient value in memory based on the current contact area. The collaborative control module sets the calculated value as the limit trigger condition of the gripping execution end and outputs the static geometric boundary coordinate system. The pose optimization module calculates the minimum overlapping area required to establish stable gripping based on the material surface roughness coefficient and static friction factor, corrects the static geometric boundary coordinate system, and locks the optimal contact surface before physical contact occurs. The collaborative control module anchors the gripping point based on the static geometric boundary coordinate system. The mechanical gripper is driven to fill the corresponding boundary according to the pre-configured topological interlocking parameters. The pressure feedback unit collects the real-time contact pressure at the end of the mechanical gripper at the moment of contact and offsets the difference between the real-time contact pressure and the normal constraint threshold. When the deviation from the normal constraint threshold is reached, a tension force compensation signal is generated. The offset monitoring module uses a visual feature tracking algorithm to monitor the axial displacement of the material. When the axial displacement exceeds the safety deviation threshold of 0.5mm to 2.0mm, the limit locking component is activated. The motion compensation unit calculates the dynamic inertial load based on the real-time acceleration of the mechanical gripper and the material mass parameters, and feeds forward to correct the driving torque. The above physical steps convert the nonlinear dynamic response process into a topological matching process in geometric space. By anchoring the pre-set static geometric boundary, the axial displacement of the material to be gripped at the moment of contact is limited, and the normal stress generated by clamping is controlled to be lower than the preset elastic limit and evenly distributed on the contact surface, thus maintaining the structural stability of the material to be gripped in the vertical working condition.

[0038] Example 2: This example addresses the structural instability of multi-specification reel materials under vertical storage conditions. A physical verification platform is constructed, integrating a 3D structured light vision camera with a spatial resolution of 0.05mm and an end-effector multidimensional force sensor with a sampling frequency of 1000Hz. Random stray light sources with superimposed illumination ranging from 300lx to 1200lx are used in the test space to simulate an industrial visual physical disturbance environment. Initial normal tilt angle offset gradients of 1, 3, and 5 degrees are set for the reel materials. The normal constraint threshold is set based on the physical balance between the minimum static friction force required for anti-slip and the maximum allowable stress for anti-rolling yield. The system solves for this threshold based on the correlation between the topological coincident volume and the preset elastic modulus. When the local missing rate of the visual point cloud increases with the increase of the initial normal tilt angle, the pose optimization module reduces the target interference depth to avoid local collisions. At the same time, the minimum coincident area constraint is amplified based on the static friction factor. Under an initial normal tilt angle of 3 degrees, the calculated optimal contact surface area is 1250mm². 2 The set normal constraint threshold is 15.5N.

[0039] A comparative sample group employing pressure feedback posterior compensation logic and an innovative sample group employing static geometric boundary control logic were established. With an initial normal tilt angle of 1 degree, the input consisted of raw visual point cloud data containing Gaussian fluctuation noise with an average amplitude of 0.4 mm. The model reconstruction module of the innovative sample group fitted a parameterized 3D mesh model and output a static geometric boundary coordinate system. The measured peak contact pressure at the moment of contact by the mechanical claw was 12.1 N, lower than the normal constraint threshold, and the axial displacement of the reel remained at 0.15 mm. Under the same conditions, the peak contact pressure of the comparative sample group reached 28.4 N, and the axial displacement reached 1.8 mm. As the initial normal tilt angle increased to 3 degrees, the local missing rate of the original point cloud increased to 12%. The innovative sample group updated the topological nesting pose according to the interference surface normal gradient variance minimization criterion. The measured peak contact pressure at the moment of contact was 15.2 N, and the axial displacement remained at 0.25 mm. The axial displacement of the comparative sample group suddenly increased to 4.5 mm, resulting in physical slippage.

[0040] Under a 5-degree initial normal tilt angle, the nonlinear response boundary characteristics of the static geometric boundary coordinate system were verified. The mechanical gripper was driven to cross the static geometric boundary coordinate system and interfere downward by 2.0 mm. The pressure sequence collected by the end-effector multidimensional force sensor showed a nonlinear surge, with the contact pressure reaching 45.0 N within 10 ms, breaking through the theoretical yield boundary. The normal strain rate at the material edge became 4.5 times that of the previous sampling period. This nonlinear inflection point data provides objective verification evidence, indicating that crossing the set boundary causes the contact surface to enter the plastic yield zone from the elastic deformation zone. After restoring the constraints of the static geometric boundary coordinate system, the measured peak contact pressure of the sample group under the 5-degree working condition of this invention returned to 18.6 N, and the residual deformation of the reel was less than 0.05 mm. The gradient verification data and the nonlinear inflection point response phenomenon confirm that the gripping dynamic response process is converted into a topological matching process in geometric space, which limits the axial displacement of the reel material to be gripped at the moment of contact, so that the normal stress generated by clamping is lower than the preset elastic limit and is evenly distributed on the contact surface, thus controlling the structural stability of multi-specification thin-walled reel materials under vertical placement conditions with stray light source disturbance.

[0041] Example 3: In the high-speed automated transfer of vertically placed reels of materials on a semiconductor chip packaging and testing production line, the mechanical gripper is subjected to alternating dynamic acceleration. The production environment is superimposed with stray light sources with random fluctuations. The coupling of dynamic inertial load and background light fluctuations interferes with the feature matching logic of the visual feature tracking algorithm, causing the loss of physical edge feature tracking of the reel, resulting in failure of axial displacement monitoring and false triggering of the limit locking component. The offset monitoring module acquires the real-time visual point cloud sequence and simultaneously measures the ambient light gradient. The offset monitoring module calculates the local surface normal vector of the reel edge, projects the real-time visual point cloud sequence onto the normal plane, and applies the illumination-independent gradient threshold operator to extract the geometric contour. The offset monitoring module constructs a sliding time window containing 10 sampling periods, calculates the optical flow vector of the geometric contour of adjacent frames, and solves and outputs the real-time axial displacement.

[0042] The collaborative control module extracts the joint angular acceleration of the robotic gripper, constructs the dynamic inertia tensor matrix of the material to be gripped from the reel based on a parametric 3D mesh model, and the motion compensation unit calculates the feedforward compensation force according to the feedforward compensation force formula. ,in, The feedforward compensation force is given by m, where m is the mass of the material to be grabbed from the reel. For the real-time spatial acceleration of the robotic gripper's end effector, To determine the vibration damping coefficient, the motion compensation unit initiates the no-load step response program of the mechanical gripper, reads the steady-state establishment time, and determines the vibration damping coefficient by calculating the ratio of this steady-state establishment time to the preset theoretical response time. The collaborative control module superimposes the feedforward compensation force onto the driving torque of the mechanical gripper. This system maintains the tracking stability of the geometric profile in the stray light source fluctuation environment by limiting the dynamic inertial deviation of the material to be gripped on the reel, ensuring the structural integrity and clamping reliability of the material to be gripped on the reel during dynamic transfer.

[0043] Example 4: In the pre-deployment scenario of introducing new specifications of reel materials into a semiconductor chip packaging and testing production line, the differences in physical properties of multiple specifications of materials require the system to pre-establish a static mapping relationship between visual features and mechanical parameters. The calibration module controls a 3D structured light vision camera to scan the surface of a standard sample and outputs benchmark point cloud data. At this time, the underlying operator calibration program is activated simultaneously, continuously acquiring 10 frames of standard grayscale images under different background illuminations. The arithmetic mean of the local neighborhood of each pixel's grayscale value (3 rows and 3 columns) is calculated and subtracted from the pixel's current grayscale value. The maximum value of this difference matrix sequence is extracted and solidified into a read-only register as an anti-illuminance measure. The absolute segmentation threshold of the interference forms a comparison instruction that can be directly called by the subsequent hardware. The feature extraction operator extracts the center aperture feature and end face flatness data of the reference point cloud data to construct the material specification feature vector. The test module controls the high-precision material testing machine to apply an incremental normal load to the standard sample and records the linear stage slope of the load displacement curve. The calculation module calculates the calibrated elastic modulus based on the linear stage slope, the cross-sectional area of ​​the standard sample, and the initial thickness. The data mapping module associates the material specification feature vector with the calibrated elastic modulus and writes it into the preset workpiece library. This offline data filling process outputs the prior physical baseline for the new specification reel material.

[0044] Under the initial calibration conditions of introducing different batches of mechanical grippers into the on-site mechanical transmission chain, the force measuring component drives the mechanical grippers to clamp a calibration reel of known mass with progressively set normal initial contact forces. The drive unit, in conjunction with the mechanical grippers, generates a vertically upward step motion. The offset monitoring module simultaneously calculates the axial slippage of the calibration reel during the motion. The optimization unit extracts the minimum normal initial contact force at which no axial slippage occurs, and applies the formula... Calculate the static friction factor of the surface, where, Let m be the static friction factor of the surface, m be the mass of the calibration reel, and g be the acceleration due to gravity. For the acceleration of the step motion, To minimize the initial normal contact force, the collaborative control module stores the surface static friction factor in a local register. The system then outputs an initial calibration dataset containing material specification feature vectors, calibrated elastic modulus, and surface static friction factor through offline material feature storage and on-site dynamic parameter self-calibration steps, defining the initial physical quantization boundary of the dynamic force matching calculation input.

[0045] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A robot grasping control system for irregular-shaped workpieces based on visual guidance and dynamic force matching, characterized in that, The system includes: The image acquisition module acquires visual point cloud data of the material to be grabbed from the reel; The model reconstruction module reconstructs a parametric 3D mesh model of the reel material to be grasped based on visual point cloud data, and retrieves a preset 3D envelope model of the mechanical claw's kinematics. The pose optimization module, within the computer-aided design logic framework, uses Boolean interference intersection operations between a parametric 3D mesh model and a 3D envelope model of the mechanical gripper's kinematics to retrieve the geometrically constrained nested poses that maximize the topological overlap volume without exceeding the theoretical yield boundary of the material to be gripped. Based on the contact surface area corresponding to the topological overlap volume and the preset elastic modulus of the material to be gripped, it calculates and generates the normal constraint threshold and outputs the static geometric boundary coordinate system. The collaborative control module anchors the gripping point based on the static geometric boundary coordinate system and performs closed-loop offsetting between the real-time contact pressure at the end of the mechanical claw and the normal constraint threshold. By driving the mechanical claw to grip the reel material to be gripped, the dynamic response process of gripping is converted into a topological matching process in geometric space, which is used to limit the axial displacement of the reel material to be gripped at the moment of contact.

2. The robotized system for picking up a profiled workpiece based on visual guidance and dynamic force matching according to claim 1, characterized in that, When the pose optimization module searches for the optimal contact surface, it applies the criterion of minimizing the variance of the normal gradient of the interference surface. The collaborative control module adjusts the mechanical claw according to the optimal contact surface to make the clamping stress evenly distributed on the surface of the reel material to be gripped, so as to control the normal stress generated by clamping to be lower than the preset elastic limit of the reel material to be gripped, and maintain the structural stability of the reel material to be gripped under vertical conditions.

3. The robotized system for picking up a profiled workpiece based on visual guidance and dynamic force matching according to claim 1, characterized in that, The model reconstruction module extracts the material specification feature vector based on the center aperture characteristics and end face flatness data of the material to be grasped from the reel material, and matches the corresponding elastic modulus parameters and geometric structure parameters from the preset workpiece library to support the dynamic fitting of the parametric three-dimensional mesh model.

4. The robotized system for picking up irregular shaped workpieces based on visual guidance and dynamic force matching according to claim 1, characterized in that, When the pose optimization module retrieves the nested pose of geometric constraints, it calculates the minimum overlapping area required to establish a stable clamping in the virtual geometric space using the material surface roughness coefficient and static friction factor, and corrects the static geometric boundary coordinate system based on the minimum overlapping area.

5. A robot gripping and control system for irregularly shaped workpieces based on vision guidance and dynamic force matching according to claim 1, characterized in that, The pose optimization module defines the topological interference volume by solving the intersection of the parametric 3D mesh model and the 3D envelope model of the mechanical gripper's kinematics in 3D space. Topological interference volume The calculation method is as follows: ,in, For topological interference volume, The spatial region occupied by the parametric 3D mesh model. The spatial region occupied by the three-dimensional envelope model of the mechanical claw's kinematics.

6. The robot gripping and control system for irregularly shaped workpieces based on vision guidance and dynamic force matching according to claim 1, characterized in that, The collaborative control module includes a pressure feedback unit. After the mechanical gripper comes into contact with the material to be gripped from the reel, the pressure feedback unit collects the real-time contact pressure at the end of the mechanical gripper and performs a difference calculation between the real-time contact pressure and the normal constraint threshold. When the real-time contact pressure deviates from the normal constraint threshold, a tension force compensation signal is generated.

7. A robot gripping and control system for irregularly shaped workpieces based on vision guidance and dynamic force matching according to claim 1, characterized in that, The system also includes an offset monitoring module, which uses a visual feature tracking algorithm to monitor the axial displacement of the material to be gripped on the reel during the movement process. When the axial displacement exceeds the preset safety deviation threshold, the collaborative control module drives the mechanical claw to work together with the limit locking component. The safety deviation threshold ranges from 0.5mm to 2.0mm.

8. The robot gripping and control system for irregularly shaped workpieces based on vision guidance and dynamic force matching according to claim 1, characterized in that, The collaborative control module also includes a motion compensation unit. The motion compensation unit calculates the dynamic inertial load based on the real-time acceleration of the mechanical gripper and the mass parameters of the material to be gripped from the reel. Based on the dynamic inertial load, it performs feedforward correction on the driving torque of the mechanical gripper to maintain the dynamic balance during the gripping process.

9. A robot gripping and control system for irregularly shaped workpieces based on vision guidance and dynamic force matching according to claim 1, characterized in that, The image acquisition module includes a multi-angle industrial camera group and a supplementary lighting unit. Before acquiring visual point cloud data, the image acquisition module uses a feature point matching algorithm to filter out environmental reflection noise from the surface of the material to be grasped on the reel, so as to improve the reconstruction accuracy of the parametric 3D mesh model.

10. A robot gripping and control system for irregularly shaped workpieces based on vision guidance and dynamic force matching according to claim 1, characterized in that, The system converts the mechanical gripper into a parameterized constraint entity within a computer-aided design logic framework. Before the physical gripping action is initiated, the geometric topology of the parameterized three-dimensional mesh model is nested, enabling adaptive anti-offset control of vertically placed reel materials of various specifications in the semiconductor packaging and testing station.

Images