Car body side grabbing device and method

Abstract

The invention discloses a high-precision and high-load automobile body side grabbing system and method, and belongs to the field of automobile manufacturing automation. The system comprises a first transfer robot and a second transfer robot which are arranged side by side; the overall cooperative gripper is connected to the robot and is composed of a first gripper frame and a second gripper frame which are modularized, and the distance between gripping points of the first gripper frame and the second gripper frame can be automatically adjusted through a ball screw structure; the binocular camera group is arranged on the gripper; and the control system is in communication connection with the robot and the camera group. Through cooperation of the double robots, the flexible adjustable gripper and double dynamic visual correction, the problems that in the prior art, the load capacity and precision are difficult to consider, flexibility is poor, and dynamic errors cannot be compensated are solved, and the stability, precision and efficiency of automatic assembly of heavy parts are remarkably improved.

Description

Technical Field

[0001] This invention discloses a vehicle body side panel gripping device and method, belonging to the field of vehicle body panel welding technology. Background Technology

[0002] This invention relates to the field of automotive manufacturing automation, and more specifically, to a high-precision, high-load gripping and handling system for the side panel assembly of a body-in-white.

[0003] In automotive welding production lines, the body side panel, as a key core component of the body-in-white, directly determines the assembly quality and appearance of the entire vehicle through the precision of its gripping, handling, and assembly positioning. It is a crucial technological link in automated production. However, existing traditional technical solutions generally suffer from one or more of the following technical shortcomings when dealing with increasingly stringent production rhythms, precision requirements, and the need for flexible production of multiple vehicle models: 1. Limited Load Capacity and Accuracy of Traditional Single-Robot Systems: The weight of the vehicle side panel assembly typically exceeds 170kg, making it difficult for traditional single-robot systems to meet such high load requirements. Prolonged operation under high loads not only drastically accelerates wear on robot joints, leading to a decline in positioning accuracy over time, but also poses safety risks. To address this issue, the industry has proposed solutions such as the "integrated gripper and clamping device" disclosed in Chinese Utility Model Patent CN207372549U. This solution attempts to improve efficiency by integrating the gripper and clamping device into a single frame. However, this integrated design results in a bulky and cumbersome actuator structure, further exacerbating the robot's load burden and failing to fundamentally solve the problem. Instead, it exacerbates the conflict between load capacity and gripping accuracy.

[0004] 2. Existing gripping devices are rigid but lack flexibility, making them unable to adapt to dynamic deviations: Traditional technologies, represented by the aforementioned "integrated gripper fixture device," heavily rely on purely mechanical pre-positioned blocks and limiting locking devices for positioning. This rigid positioning design makes it lack the ability to adapt to dynamic changes during the production process. In actual working conditions, the side panels of the incoming material may have slight deformations, or their position and orientation may shift during transportation. Simultaneously, the fixture itself has installation tolerances. This type of "blind" system cannot perceive and compensate for these dynamic deviations in real time, causing errors to be directly transmitted to subsequent assembly processes, resulting in fluctuations in the final assembly accuracy.

[0005] 3. Insufficient collaborative control level of existing dual-robot systems: To address the load bottleneck of single robots, some production lines have begun to apply dual-robot systems. However, most existing dual-robot systems adopt a fixed master-slave control mode and lack advanced dynamic load balancing mechanisms. During collaborative operations, uneven grasping force or interference in motion trajectories between the two robots can easily occur, affecting not only the stability of grasping but also failing to fully realize the potential of dual-robot collaborative work.

[0006] 4. Traditional vision-based deviation correction systems suffer from response lag: Although some high-end production lines have introduced vision systems for deviation correction, traditional vision systems mostly rely on offline calibration. This means that they cannot compensate for dynamic deviations caused by workpiece deformation, fixture errors, etc., in real time within the production cycle, resulting in significant response lag. This greatly reduces the effectiveness of vision-based deviation correction and makes it difficult to meet the demands of high-precision, high-speed modern production.

[0007] In summary, existing technologies generally suffer from a series of technical bottlenecks when facing high-load automotive side panel gripping applications, including insufficient load capacity, low gripping accuracy, poor adaptability to dynamic deviations, inadequate collaborative control, and slow response speed. Therefore, the industry urgently needs a new technical solution that can effectively integrate high-load handling capacity, high-precision dynamic correction, and high-response speed to achieve stable, accurate, and efficient automated assembly of the vehicle side panel assembly. Summary of the Invention

[0008] The following technical problems are commonly found in existing automotive body side gripping systems: 1) Traditional single-robot systems have insufficient load capacity, and their integrated gripper solutions are bulky and thus affect positioning accuracy; 2) Existing gripping devices are mostly rigid structures, which cannot compensate for dynamic errors such as material orientation deviation and self-deformation in real time; 3) Existing dual-robot systems have insufficient collaborative control, which easily leads to uneven gripping force or motion interference; 4) Traditional vision correction systems rely on offline calibration, have slow response speed, and cannot meet the production requirements of high precision and fast cycle time.

[0009] The purpose of this invention is to provide a high-precision, high-load vehicle side panel gripping system and gripping method, aiming to solve the technical bottlenecks in load capacity, gripping accuracy, dynamic adaptability and response speed, and to achieve stable, accurate and efficient automated gripping and assembly of heavy vehicle side panel assemblies.

[0010] To achieve the above objectives, the present invention provides the following technical solution: A high-precision, high-load vehicle side panel gripping system. To solve the above technical problems, the present invention provides a high-precision, high-load vehicle side panel gripping system, comprising an installation platform (5); a first transport robot (1.1) and a second transport robot (1.2), wherein the first and second transport robots are installed side-by-side on the installation platform and are connected via communication to achieve collaborative control; and an integrated collaborative gripper, which is composed of a modularly designed first gripper with a camera and a second gripper with a camera. The first gripper with a camera is mechanically connected to the first transport robot (1.1), and the second gripper with a camera is mechanically connected to the second transport robot (1.2). The system controls two transport robots in a coordinated manner to drive the overall collaborative gripper to accurately transfer the side assembly located on the panel transfer trolley (3) to the assembly fixture (4); a binocular camera group is symmetrically arranged on the top crossbeam of the overall collaborative gripper. Its function is to obtain the three-dimensional spatial pose information of the side panel of the vehicle body through stereo vision technology before the formal gripping, calculate its deviation from the theoretical position, and generate visual signals to guide the robot's movements. A control system, acting as the brain of the entire system, communicates at high speed with the two handling robots and the binocular camera array. It is responsible for receiving and processing visual guidance signals, and based on the deviation values ​​calculated from these signals, for real-time correction of the movement trajectories of the first and second handling robots.

[0011] In view of the above system, the present invention also provides the following further embodiments: Preferably, in a preferred embodiment of the present invention, the first gripper with camera is specifically structured as a first gripper frame, and the second gripper with camera is specifically structured as a second gripper frame. The first gripper frame includes a first member serving as a main beam and multiple second members perpendicular to the first member. The first member has a flange interface for connecting to the first handling robot (1.1). Similarly, the second gripper frame includes a third member serving as a main beam and multiple fourth members perpendicular to the third member. The third member has a flange interface for connecting to the second handling robot (1.2). To achieve gripping and positioning functions, each second and fourth member is connected to a gripper and a camera, wherein the gripper and the member are connected by a screw mechanism for easy fine-tuning.

[0012] Preferably, to achieve flexible adaptation to side panels of different vehicle models, the present invention further specifies that: the multiple second rods are adjustablely connected to the first rod via a ball screw structure. These ball screws are arranged along the length direction of the first rod, and by driving the ball screws, the spacing between the second rods can be precisely adjusted while maintaining their parallelism. Similarly, multiple fourth rods are also adjustablely connected to the third rod via the same ball screw structure, realizing flexible adjustment of the gripping point spacing, thereby adapting to side panels of vehicle models with different wheelbases or profile dimensions.

[0013] Preferably, to ensure the accuracy and stability of visual measurements, the binocular camera group consists of at least two sets of high-resolution industrial binocular cameras arranged symmetrically. Crucially, the baseline distance between the two cameras in each set is precisely set to 300mm, an optimized distance that balances measurement depth of field and accuracy.

[0014] Preferably, the control system incorporates an advanced visual guidance algorithm based on deep learning. This algorithm can automatically identify and lock at least six preset key feature points (such as positioning holes, edge contour features, etc.) on the side panel, and based on these feature points, accurately calculate the displacement deviation of the panel relative to the preset reference position along the X, Y, and Z coordinate axes, as well as the rotational deviation around these three axes, thereby completing a comprehensive deviation calculation of the six degrees of freedom of the panel's spatial pose.

[0015] Corresponding to claim 6, to meet production cycle requirements, the visual guidance algorithm is highly optimized, and the time required to complete a single six-degree-of-freedom deviation calculation is no more than 20 milliseconds (ms). Thanks to the high-precision hardware and algorithm, the overall positioning accuracy of the entire visual guidance system can reach ±0.25 millimeters (mm), which is far superior to traditional solutions.

[0016] Preferably, to ensure real-time data transmission and high response speed, the communication between the control system and the two handling robots preferably adopts the PROFINET industrial Ethernet protocol. This protocol provides a communication rate of up to 100 Mbps and ensures that the signal transmission delay from the completion of deviation calculation to the start of the robot's corrective action is no more than 5 milliseconds (ms), thereby greatly shortening the overall system response time and ensuring the timeliness of the corrective action.

[0017] The present invention also provides a high-precision and high-load body side panel grasping method supporting the above system. Through precise step and parameter control, the technical solution is implemented. The method specifically includes the following steps: a. Visual scanning and data processing stage: First, coordinate and control two handling robots to drive the overall collaborative gripper to move into an optimized visual scanning range of 500 mm to 1500 mm away from the target side panel assembly. Then, start the binocular camera group and control it to take three consecutive photos of the side panel assembly according to a preset Z-shaped scanning trajectory. To improve the measurement stability, the position data obtained from the three measurements are averaged arithmetically as the final raw deviation data. b. Deviation calculation and primary trajectory correction stage: After the control system receives the final raw deviation data, it immediately processes it to calculate the precise displacement and angular deviation values of the side panel assembly in six degrees of freedom. Subsequently, through the PROFINET industrial Ethernet, these deviation data are transmitted at high speed to the controllers of the two handling robots, and the robot controllers automatically generate a corrected motion trajectory accordingly. c. Precise grasping and torque balance stage: The two handling robots move collaboratively strictly according to the newly generated corrected trajectory, driving multiple grasping points on the overall collaborative gripper to precisely fit the preset grasping points on the side panel assembly. During the clamping process, the load distribution system in the system will monitor and adjust the output torques of the two robots in real time to ensure that the grasping forces on both sides are uniform and avoid panel deformation caused by uneven stress. d. Dynamic monitoring and secondary trajectory correction stage: During the process of transporting the side panel assembly from the transfer cart (3) to the merging fixture (4), the binocular camera group switches to the real-time monitoring mode to continuously monitor the spatial attitude of the side panel assembly. Once it is detected that the attitude change deviation caused by factors such as movement and vibration exceeds the preset threshold, the system will immediately give feedback and trigger the robot to perform a second real-time trajectory correction. e. Precise placing stage: Finally, the two handling robots place the side panel assembly smoothly and precisely at the specified positioning position of the merging fixture (4) according to the final trajectory after one or two corrections.

[0018] Preferably, to further elaborate on the reliability of the deviation calculation in step b, the present invention further explains that the calculation of the six-degree-of-freedom deviation is completed through a vision-guided algorithm based on deep learning. The core advantage of this algorithm is that it can automatically identify at least 6 key feature points on the side panel assembly through a pre-trained neural network model and has strong anti-interference ability, which can effectively exclude visual interference caused by factors such as oil stains, reflections, or slight scratches on the panel surface, thus ensuring the accuracy of the three-dimensional coordinate calculation.

[0019] Preferably, for the secondary correction triggering condition in step d, the present invention further specifies that the preset deviation threshold for triggering the secondary trajectory correction is ≥0.5mm. This means that during the handling process, any disturbance that causes the plate's posture to deviate from the target trajectory by more than 0.5mm will be captured by the system and corrected immediately, ensuring the final placement accuracy.

[0020] The beneficial effects of this invention are as follows: Compared with the prior art, this invention is not a simple improvement on a single technical point, but rather achieves a series of significant beneficial effects through deep integration and system innovation of dual-robot collaborative control, high-precision dynamic vision guidance, flexible mechanical structure, and high-speed industrial communication. These effects are specifically reflected in the following aspects: First, this invention fundamentally solves the contradiction between "high load" and "high stability" in heavy component handling. By creatively employing a dual-robot collaborative system and combining it with an advanced dynamic load balancing control algorithm, the weight of the vehicle body side assembly, exceeding 170kg, can be evenly distributed between the two robots. This not only completely avoids the accelerated joint wear, shortened service life, and potential safety risks caused by severe overloading in traditional single-robot systems, but also ensures stable posture during high-speed handling through collaborative motion planning, significantly improving the system's reliability and stability under long-term, high-intensity operation.

[0021] Secondly, this invention achieves a qualitative leap in "high precision" and "dynamic correction" capabilities. It integrates a high-response binocular vision system and a deep learning-based intelligent algorithm, enabling precise calculation of the workpiece's six-degree-of-freedom spatial pose within 20 milliseconds. Its core innovation lies in establishing a dual closed-loop control mechanism of "pre-grab correction" and "secondary correction during handling." This not only compensates for incoming material positional deviations before grasping but also monitors and corrects attitude deviations caused by vibration or inertia in real time during movement, achieving full-process, real-time compensation for various dynamic errors. This results in a final placement accuracy of ±0.25mm, unmatched by traditional "blind" systems relying on mechanical positioning or offline calibration.

[0022] Furthermore, this invention demonstrates significant advantages in terms of "high adaptability" and "production flexibility." The gripper frame innovatively employs a modular design and introduces an adjustable structure with ball screw-driven lever spacing. This design allows the gripping point position to be quickly and automatically adjusted according to the side dimensions of different vehicle models, without requiring time-consuming and labor-intensive hardware replacements or complex mechanical adjustments. This greatly enhances the production line's flexibility and rapid changeover efficiency in handling the current automotive industry's multi-model, small-batch, mixed-line production model.

[0023] Finally, this invention achieves new heights in both "high efficiency" and "industrial applicability." Thanks to efficient vision algorithms and high-speed industrial Ethernet communication based on the PROFINET protocol, the response time of the entire decision-making and control chain—from visual scanning, deviation calculation, trajectory correction to robot execution—is compressed to its limit. This ensures that complex grasping operations involving multiple dynamic corrections can still be completed efficiently within the stringent automotive production cycle, thus effectively guaranteeing the automation level and production efficiency of the entire production line without sacrificing, and even improving, assembly quality, making it highly valuable for industrial application. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments disclosed in this invention, the accompanying drawings of the embodiments will be briefly described below. These drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention.

[0025] Figure 1 This is a schematic diagram of the overall structure of the vehicle side panel gripping system provided in one embodiment of the present invention.

[0026] Figure 2 This is a three-dimensional schematic diagram of the first camera-equipped gripper of the dual-robot overall collaborative system in an embodiment of the present invention.

[0027] Figure 3 This is a three-dimensional schematic diagram of the second camera-equipped gripper of the dual-robot overall collaborative system in an embodiment of the present invention.

[0028] Figure 4 This is a schematic diagram of the overall collaborative gripper in the state of not gripping a board in an embodiment of the present invention.

[0029] Figure 5 This is a schematic diagram of the overall collaborative gripper in the state of gripping the side panel in an embodiment of the present invention.

[0030] Figure 6 This is a load characteristic curve output by the dual-robot collaborative system in a practical application according to an embodiment of the present invention, used to intuitively demonstrate the high load capacity and load stability of the present invention. Explanation of reference numerals in the attached drawings: 1-Dual robot overall collaborative system, 1.1-First handling robot, 1.2-Second handling robot, 2-Overall collaborative gripper, 2.1-First gripper with camera, 2.2-Second gripper with camera, 3-Panel transfer trolley, 4-Assembly fixture, 5-Installation platform, 6-Slide table, 7-Binocular camera group, 8-Side assembly. Detailed Implementation

[0031] The technical solutions (including preferred technical solutions) of the present invention will be further described in detail below with reference to the accompanying drawings and by way of listing some optional embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. It should be understood that the embodiments described herein are merely for explaining this invention and do not constitute any limitation on the scope of protection of this invention. Based on the content disclosed in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection claimed by this invention.

[0033] This invention provides a highly intelligent, flexible, high-precision, high-load vehicle side panel gripping system. Please refer to [link / reference]. Figure 1-5 The typical application scenario of this system on the automotive welding production line is as follows: the body side panel assembly (usually a large thin-walled stamped part weighing more than 170kg) transported from the upstream process to the designated station via the sheet metal transfer trolley 3 is precisely and stably gripped and transported to the downstream assembly fixture 4 for precise positioning and assembly welding with the body floor, roof and other components.

[0034] The physical entity of the system is primarily constructed on a precision-machined and aged mounting platform 5, possessing extremely high rigidity and stability. The core components of the system are integrated onto this platform, mainly including: a dual-robot collaborative system 1, a collaborative gripper 2, a high-precision binocular camera group 7, and a central control system (not shown separately in the diagram) serving as the system's "brain." In some flexible production line layouts, the assembly fixture 4 itself can also be moved via a high-precision servo slide 6 to adapt to adjustments in the production of different vehicle models or process flows.

[0035] The cornerstone of this invention is the innovative use of a dual-robot collaborative workstation consisting of a first handling robot 1.1 and a second handling robot (1.2). These two six-axis industrial robots (e.g., models with matched arm span and load capacity) are mounted side-by-side on the mounting platform 5 in a strictly symmetrical manner. The reason for abandoning the traditional single heavy-duty robot solution is based on the following in-depth considerations: Increased load capacity and ensured safety margin: When a single robot faces a load exceeding 170kg, even if its nominal load capacity is sufficient, it is often operating at the edge of its performance limit. Prolonged operation under these conditions not only drastically accelerates the wear of core components such as reducers and bearings, causing the robot's repeatability accuracy to deteriorate over time, but also poses a risk of shutdown or even safety accidents due to overload or dynamic impacts. Adopting a dual-robot system distributes the total load almost evenly between the two robots, ensuring that each robot operates within its rated load range, significantly extending the equipment's lifespan and guaranteeing absolute operational stability.

[0036] Improved motion stability: For large, thin-walled components like side panels, single-point or single-arm support is prone to swaying, vibration, and even torsional deformation during high-speed handling. Dual robots working together to support and transport from both sides create a wide and stable support base, significantly suppressing workpiece attitude changes during high-speed start-stop and turning, laying a solid foundation for subsequent high-precision placement.

[0037] Collaborative control: The two robots do not work independently, but communicate at the low level via deterministic high-speed industrial Ethernet (preferably PROFINET protocol), with unified kinematic and dynamic planning performed by the central control system. The control system can achieve millisecond-level program synchronization and motion trajectory synchronization between the two robots, ensuring highly consistent movements, resembling a coordinated and unified organic whole.

[0038] The integrated collaborative gripper 2 is the core embodiment of structural innovation and functional integration in this invention. It completely overturns the traditional integrated, rigid, and bulky gripper design concept, and its ingenuity is reflected in the following aspects: like Figure 4 As shown, the overall collaborative gripper 2 is structurally composed of two independent and functionally symmetrical sub-units—the first camera-equipped gripper 2.1 and the second camera-equipped gripper 2.2—which are logically integrated through collaborative control. The first camera-equipped gripper 2.1 is rigidly connected to the sixth axis of the first handling robot 1.1 via its end flange, while the second camera-equipped gripper 2.2 is connected to the second handling robot 1.2.

[0039] Each sub-unit (taking 2.1 as an example) has an independent gripper frame 2.1.1, which is made of lightweight, high-strength alloy steel pipe and designed through topology optimization using finite element analysis (FEA). This design minimizes the weight of the gripper itself while ensuring sufficient rigidity and strength, thereby reducing the robot's ineffective load and improving the system's dynamic response performance and energy efficiency.

[0040] This is one of the core innovations of this invention in terms of mechanical structure, aiming to solve the industry pain point of multi-model mixed-line production. Please refer again. Figure 4 and Figure 5 The first gripper frame includes a first member that serves as the main load-bearing beam and multiple functional second members arranged perpendicular to the first member.

[0041] The key is that these second links are not fixed. Inside or on the side of the first link, along its length, at least one high-precision ball screw driven by an independent servo motor and a matching heavy-duty precision linear guide for load bearing and guidance are precisely installed. The root of each second link is connected to the linear guide via a slider and engages with the ball screw via a high-precision screw-nut pair.

[0042] When the production line needs to switch from model A to model B, the operator simply selects the corresponding model program in the host computer (HMI). The central control system immediately retrieves the side panel digital model parameters of model B and calculates the new spacing required for each gripping point. Subsequently, control commands are sent to each servo motor, precisely driving the ball screw to rotate by the specified angle. The rotational motion of the ball screw is converted into the linear displacement of the second rod through the screw-nut pair, allowing it to move smoothly and accurately to the new preset position under the guidance of the linear guide. Thanks to the use of high-precision servo closed-loop control, its positioning accuracy can reach ±0.03mm. The entire adjustment process is fully automated and can be completed within seconds.

[0043] The third and fourth links of the second camera-equipped gripper 2.2 also employ the same symmetrical design. This innovative design allows the gripper to automatically, quickly, and precisely adjust its gripping span and gripping point distribution to perfectly adapt to the side panels of various vehicle models with different wheelbases, profile dimensions, and gripping point layouts. This achieves true flexible production, significantly reduces changeover time, and improves the overall uptime of the production line.

[0044] At the end of each adjustable second and fourth link, a gripping actuator (e.g., a lightweight, high-strength pneumatic gripper controlled by a solenoid valve) and an auxiliary camera for precise local positioning are connected via a lead screw structure. This lead screw connection facilitates precise fine-tuning of the gripper's height in the Z-axis direction during initial installation and commissioning.

[0045] The high-precision binocular camera group 7 is the hardware foundation for this system's active perception and intelligent decision-making. It consists of two identical industrial-grade high-resolution binocular cameras, mounted symmetrically on the crossbeam atop the overall collaborative gripper 2. Its hardware selection and configuration have been thoroughly optimized. Hardware Specifications: Utilizing an industrial-grade CMOS image sensor with a physical resolution of 2592×1944 pixels, ensuring the capture of minute details on side panel components. A frame rate of up to 30fps meets the dynamic tracking requirements of high-speed motion. An ultra-wide dynamic range (WDR) of no less than 120dB is crucial for handling the complex and varied lighting environments in automotive welding workshops (e.g., strong arc welding light, equipment shadows, and drastic changes in ambient light levels between 500 lux and 10000 lux), ensuring clear, unexposed or underexposed images under various lighting conditions.

[0046] Optical Configuration: The physical distance between the two lenses of each binocular camera group, i.e., the baseline, is precisely calibrated and fixed at 300mm. This distance is an optimized result that achieves the best balance between measurement accuracy and measurement depth of field (effective working range). A longer baseline helps to improve the resolution of depth measurement, thereby improving the accuracy of 3D coordinate calculation.

[0047] Installation and Calibration: The entire camera assembly underwent high-precision calibration to obtain its accurate internal parameters (such as focal length, principal point coordinates, and distortion coefficients) and external parameters (i.e., the precise transformation matrix between the camera coordinate system and the robot's end effector flange coordinate system). This "eye-in-hand" configuration allows visual measurement results to be directly and accurately converted into the robot's working coordinate system, providing a reliable data foundation for subsequent trajectory correction.

[0048] The central control system is the command center of the entire system, typically composed of a high-performance industrial PC (IPC) or programmable logic controller (PLC). It integrates all functions, including motion control, vision processing, logic control, and human-machine interaction. Its main responsibilities include: Cooperative motion planning: Responsible for receiving the overall operation instructions and breaking them down into cooperative motion paths for the two handling robots, ensuring the synchronization, coordination and collision-free interference of their movements.

[0049] Visual data processing: The core visual guidance algorithm is run to process the image data collected by the binocular camera group (7) and complete complex calculations such as feature point recognition, three-dimensional reconstruction, and pose deviation calculation.

[0050] Real-time closed-loop control: The calculated deviation data is sent to the robot controller via high-speed industrial Ethernet (preferably PROFINET protocol, whose deterministic data exchange and extremely low latency (≤5ms) are key to achieving high response speed) to realize real-time correction of the robot's motion trajectory.

[0051] Torque balance control: Real-time monitoring and adjustment of the joint torque of the two robots to achieve active force balance.

[0052] Logic and safety control: Manage the logical sequence of the entire workflow and interact with other equipment on the production line (such as conveyor trolleys, assembly fixtures, safety light curtains, etc.) via I / O signals to ensure smooth operation and the safety of personnel and equipment.

[0053] The grasping method protected by this invention is an intelligent workflow that deeply integrates the aforementioned precision hardware system with advanced algorithm software. It completely revolutionizes the traditional "teach-and-reproduce" mode into a precision workflow that integrates system self-checking, multi-vision closed-loop control, dynamic path planning, and intelligent torque balancing, aiming to achieve ultimate precision control over high-load workpieces.

[0054] S1: At the start of each production shift, or upon restarting the system after a prolonged shutdown, the control system will forcibly execute a comprehensive initialization and self-check procedure. This stage is fundamental to ensuring the accuracy of all subsequent operations.

[0055] Robot Zero Point Return and Coordinate System Calibration: The first handling robot 1.1 and the second handling robot 1.2 will automatically execute the procedure to return to the mechanical zero point. By touching the built-in zero-position sensor, the robot controller will recalibrate the encoders of each joint, thereby establishing a precise robot base coordinate system and tool coordinate system without cumulative error.

[0056] Online Automatic Calibration of Vision Systems: Unlike traditional vision systems that require "one-time calibration for long-term use," this invention introduces the concept of online automatic calibration. A high-precision checkerboard or dot array calibration board is installed at a specific location on the workstation. During the self-test process, dual robots drive the gripper, causing the binocular camera group 7 to photograph the calibration board from multiple different angles and distances. The control system automatically runs the calibration algorithm, recalculating and fine-tuning the camera's intrinsic and extrinsic parameters. This process can compensate in real time for minor mechanical deformations caused by long-term equipment operation, or accuracy drift caused by thermal expansion and contraction of optical components due to changes in ambient temperature, ensuring that the results of every vision measurement remain at the highest accuracy level.

[0057] S2, the core objective of this stage is to completely and accurately compensate for the static positional deviation of the side assembly before the robot makes any physical contact with the workpiece.

[0058] Optimized interval scanning: After the operation begins, when the panel conveyor trolley 3 transports the side panel assembly to the workstation and sends out the "in position" signal, the dual robots work together to drive the overall collaborative gripper 2 to move smoothly to the visual scanning work area 500mm to 1500mm away from the panel.

[0059] "Z"-shaped depth scanning and data fusion: To construct a complete and blind-spot-free 3D point cloud model, the camera array does not perform a simple single image, but instead performs three consecutive scans along a pre-set "Z"-shaped trajectory covering all key feature areas of the side assembly. This scanning path allows the cameras to capture feature points from multiple different advantageous perspectives (e.g., top view, left front, right rear), effectively overcoming problems such as local reflections caused by changes in the curvature of the metal plate surface, shadows at specific angles, or local occlusions caused by structural features under a single perspective. The control system performs data fusion processing on the millions of 3D coordinate points acquired from these three scans, using algorithms such as "Iterative Closest Point" (ICP) for registration, and then performing a weighted average on the registered point cloud. This greatly filters out random measurement noise introduced by factors such as ambient light fluctuations, airborne dust reflections, and minor equipment vibrations, obtaining a set of final deviation raw data with extremely high signal-to-noise ratio and statistical reliability.

[0060] The deep learning-based six-DOF deviation analysis: The control system utilizes its core visual guidance algorithm to analyze the raw deviation data. This algorithm embeds a pre-trained convolutional neural network (CNN) model trained through supervised learning on a dataset containing tens of thousands of vehicle side panel images from different models, under varying lighting conditions, and in different postures. This model can accurately identify at least six key feature points on the panel with extremely high robustness. These feature points can be geometrically well-defined positioning hole centers, specific tangent points of contour lines, apexes of wheel arch curves, or even the curvature centers of specific areas. The algorithm's advancement lies in its ability to "understand" the context of the features, effectively resisting visual interference from oil stains, scratches, water stains, and strong reflections on the panel surface under real-world conditions.

[0061] After identifying feature points, the system utilizes stereo matching and triangulation principles of binocular vision, combined with precisely calibrated camera parameters, to calculate the accurate 3D spatial coordinates of each feature point. Finally, through efficient coordinate system transformation algorithms (such as Singular Value Decomposition (SVD) or quaternion methods), the measured actual feature point cloud is optimally fitted and registered with the theoretical feature point cloud extracted from the CAD model. The registration result is a 4x4 homogeneous transformation matrix describing the current side panel assembly relative to its theoretical ideal position. From this matrix, the translational deviations (ΔX, ΔY, ΔZ) along the X, Y, and Z axes and the rotational deviations (Euler angles or quaternion representations ΔRx, ΔRy, ΔRz) about these three axes can be directly analyzed. This completes the comprehensive and high-precision calculation of the workpiece's six degrees of freedom deviations. The entire process (from taking the picture to outputting the deviation matrix) is strictly controlled to within 20 milliseconds.

[0062] Global dynamic trajectory generation: The calculated six-degree-of-freedom deviation matrix is ​​sent to the dual-robot controller in real time via PROFINET industrial Ethernet (signal delay ≤5ms). Unlike traditional methods that simply perform linear compensation on the endpoint coordinates of the path, the robot controller uses this deviation transformation matrix as a coordinate transformation operator, applying it to every path point along the entire preset grasping path. This means that the robot will dynamically generate a completely new, smooth trajectory that perfectly matches the actual pose of the workpiece, ensuring that the relative posture of the gripper and workpiece remains optimal throughout the entire process from initial approach to final contact.

[0063] S3, High-precision trajectory synchronous execution: Dual robots 1.1 and 1.2 move strictly according to the new trajectory generated in the second stage in master-slave synchronous or more advanced distributed cooperative control mode.

[0064] Active torque equalization control: During the gripper's contact with the workpiece, the clamping process, and the subsequent lifting phase, the control system enters an active torque equalization closed-loop control mode. It reads and compares the motor current feedback values ​​of all six joints of both robots in real time at a millisecond-level sampling frequency (this value is proportional to the joint output torque). Once an imbalance in the load torque corresponding to the two robots is detected (for example, a slight shift in the workpiece's center of gravity causing one robot to "feel heavier"), the main controller immediately generates fine-tuning instructions through a PID or more complex adaptive fuzzy controller. These instructions dynamically and differentially adjust the movement speed and acceleration of the two robots, achieving precise force-sensing control similar to two skilled athletes sensing and matching each other's force during a tug-of-war. Figure 3 The actual load curves shown are almost completely overlapping and have minimal fluctuations. This ensures that the resultant force applied to the side panel assembly is evenly distributed, fundamentally eliminating the risk of elastic deformation, stress concentration, or even permanent plastic deformation of thin-walled panels due to uneven force distribution.

[0065] S4, this stage is the revolutionary innovation of this invention that distinguishes it from all traditional static visual positioning schemes. It extends the precision control from the "static" before the operation to the "dynamic" during the operation.

[0066] High-speed visual tracking and state prediction: During the collaborative transport of the side panel assembly by the two robots, as they fly at high speed to the assembly fixture 4, the binocular camera group 7 switches to high-speed tracking mode. At this time, the visual algorithm no longer performs time-consuming global feature search, but instead utilizes state estimation algorithms such as the Kalman filter. Based on the feature point positions identified in the previous frame and the robot's kinematic model, the algorithm predicts the most likely location of the feature points in the current frame and defines a very small region of interest (ROI) at this location. The system only needs to perform feature matching within this very small ROI, greatly improving processing speed and achieving real-time, stable tracking of key feature points.

[0067] Online deviation monitoring and "overlay" correction: The system continuously compares the real-time tracked 3D positions of feature points with the theoretical trajectory positions calculated based on the robot's current joint angles. If the actual posture of the workpiece deviates from the theoretical trajectory due to the robot's high-speed inertia, minor vibrations, or the robot's accumulated trajectory errors, and the deviation exceeds a preset dynamic threshold (e.g., 0.5mm), the system immediately identifies it as a dynamic deviation event requiring intervention. At this point, the system instantly calculates a tiny correction transformation matrix and, through high-speed communication, "overlays" this matrix onto the motion command currently being executed by the robot controller. This "online overlay correction" method makes trajectory adjustment smooth and seamless, allowing the robot to complete "aerial posture correction" without any pauses. This ensures that even during high-speed travel, the workpiece carried by the gripper end effector remains in the correct posture, infinitely approaching the theoretical target trajectory.

[0068] S5, after the dual protection of the above-mentioned "static first correction" and "dynamic second correction", when the two robots 1.1 and 1.2 carrying the side assembly arrive at the top of the assembly fixture 4, their pose has achieved a very high degree of matching with the positioning reference of the fixture.

[0069] Slowing down and gently placing: In the final placement stage, the robot will automatically reduce its running speed and gently and controllably align the main positioning holes (usually Φ15mm or Φ25mm process holes) on the side assembly with the corresponding cylindrical or diamond-shaped positioning pins on the fixture, and then smoothly place it down.

[0070] Multi-point verification and release: After the main locating pin enters the locating hole, the system continues to descend until all secondary locating reference surfaces (such as planes and edges) on the side panel are also fully engaged with the corresponding support blocks of the fixture. Only after confirming that all locating references are in place and the workpiece is in a completely stress-free free state will the control system issue a command to release the workpiece from all the grippers on the gripper. This "zero-stress" placement operation ensures that the workpiece will not experience any stress due to positioning errors within the fixture, creating perfect preconditions for subsequent high-quality welding.

[0071] Thus, a complete cycle of high-precision, high-load, high-flexibility, and highly intelligent vehicle side panel gripping operation has been completed.

[0072] This invention achieves several technical effects through its system architecture design and methodological implementation. These effects stem from the organic combination of various technical features, and through synergistic effects, overall performance is improved compared to existing technologies.

[0073] I. Achieved a balance between high load capacity and high operational accuracy. In the field of robotics applications, there is often a technical constraint between high-load operation and high-precision control. This invention effectively solves this problem through the following design: First, the system employs a dual-robot collaborative load distribution system. The weight of the vehicle side panel assembly, exceeding 170 kg, is distributed between the two robots, ensuring that each robot's load remains within its rated operating range. This approach reduces the wear rate of the robot's core transmission components (such as reducers and bearings), helping to maintain its inherent repeatability and positioning accuracy over the long term. Simultaneously, the wide support base formed by the two arms enhances the posture stability of large, thin-walled workpieces during high-speed handling, providing a stable prerequisite for achieving high-precision placement.

[0074] Secondly, a lightweight modular gripper design was adopted. By using lightweight, high-strength materials optimized through finite element analysis and a modular structure, the weight of the end effector itself was reduced. This design, combined with dual-robot load sharing, further reduced the total load that the robot needs to bear (the sum of the workpiece weight and the gripper weight). The lower total load allows the robot's dynamic performance to be better utilized, enabling it to complete actions with higher acceleration while suppressing excessive vibration or attitude errors.

[0075] Finally, active torque equalization control was implemented. During the gripping and transporting of the workpiece, the system uses high-frequency torque feedback for closed-loop adjustment to ensure the uniformity of the gripping force on both sides. This avoids the risk of elastic deformation of thin-walled workpieces due to uneven force distribution, ensures the stability of the center of mass of the entire motion system, and improves the smoothness of the motion process.

[0076] II. Extending precision control from static positioning to full-process dynamic closed-loop guidance Traditional vision positioning systems are typically limited to a single positioning operation before the start of the task, with the motion process mostly involving open-loop control. This invention, through a dual correction mechanism, integrates high-precision vision control throughout the entire task process.

[0077] The first layer of correction is a highly reliable static correction before grasping. A static pose calculation system is constructed using zigzag scanning, multi-frame data fusion, and a deep learning-based algorithm. This system can accurately calculate the initial six-degree-of-freedom deviation of the workpiece and, through the robustness of the deep learning model, reduces the interference of surface conditions such as oil stains, reflections, and scratches on the recognition results. This solves the static initial error caused by factors such as inconsistent incoming material positions, tooling and fixture tolerances, or deformation of the sheet metal itself.

[0078] The second layer of correction is a high-response dynamic correction during the handling process. During high-speed workpiece handling, the system continuously acquires the workpiece's actual posture through high-speed visual tracking and online deviation monitoring. When the system detects that the actual posture deviates from the preset trajectory by more than a certain threshold (e.g., 0.5 mm) due to factors such as inertia, vibration, or robot trajectory errors, the system immediately performs smooth online trajectory correction. This mechanism enables the system to compensate for unpredictable dynamic errors generated during movement.

[0079] These two correction mechanisms, combined with high-resolution vision hardware, high-efficiency algorithms (computation time ≤ 20 milliseconds) and high-speed communication networks, constitute a dynamic closed-loop control system with high response and high error correction capabilities, ultimately achieving a high overall positioning accuracy (e.g. ±0.25 mm).

[0080] Third, it improved production flexibility and equipment utilization efficiency. To adapt to the production mode of multiple vehicle models and rapid iteration, this invention improves the system's flexibility by adopting an automated adjustable gripper mechanism based on ball screws.

[0081] This design reduces the time required for physical changeovers. When production tasks switch models, the operator only needs to select a new program in the control system, and the multiple gripping points of the gripper can automatically and precisely adjust to the new spacing and layout within a short time through the linkage of the servo motor and the ball screw. This significantly shortens the changeover time of the production line, thereby improving the overall efficiency of the equipment and its responsiveness to changes in production plans.

[0082] At the same time, this design also reduces related hardware and maintenance costs. A flexible gripper system can be adapted to various existing and future vehicle models through software configuration, eliminating the need to design, manufacture, and stock multiple sets of dedicated physical grippers for different vehicle models, thereby saving hardware investment and simplifying the management of spare parts.

[0083] IV. A stable and reliable industrial application solution has been developed. This invention provides a stable, reliable, and easy-to-maintain intelligent application solution through systematic integration.

[0084] The system possesses a certain degree of self-adaptation and self-maintenance capabilities. For example, the online automatic calibration function during the initialization phase enables the system to compensate for accuracy drift caused by long-term operation or environmental changes, reducing reliance on complex manual calibration and maintenance. The deep learning-based vision algorithm gives the system strong adaptability to changes in workpiece surface conditions, lowering the stringent requirements for production environment conditions.

[0085] The various technical features of the system work synergistically. For example, the lightweight gripper enables the dual robots to achieve high dynamic performance, which creates conditions for dynamic secondary correction during high-speed handling; while high-precision vision guidance ensures the final positioning accuracy of the flexible gripper after adjustment. This mutual support between technologies forms the basis of the reliability of the present invention.

[0086] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A vehicle side panel gripping system, characterized in that, include: Installation platform (5); a first transport robot (1.1) and a second transport robot (1.2), wherein the first and second transport robots are installed side by side on the installation platform and the first transport robot (1.1) and the second transport robot (1.2) are communicatively connected; The overall collaborative gripper includes a first camera gripper and a second camera gripper integrated through a modular structural design. The first camera gripper is connected to the first transport robot (1.1), and the second camera gripper is connected to the second transport robot (1.2). The first and second transport robots drive the overall collaborative gripper through collaborative control to transfer the side assembly located on the panel transfer trolley (3) to the assembly fixture (4). A binocular camera group is symmetrically arranged on the top of the overall collaborative gripper to obtain the three-dimensional coordinate information of the vehicle side panel before the gripping operation to calculate the deviation and generate a visual guidance signal. A control system is communicatively connected to the first and second transport robots and the binocular camera group to receive the visual guidance signal and correct the movement trajectory of the first and second transport robots in real time according to the signal.

2. The vehicle side panel gripping system according to claim 1, characterized in that, The first gripper with camera has a first gripper frame, and the second gripper with camera has a second gripper frame. The first gripper frame includes a first rod and a plurality of second rods perpendicular to the first rod. The first rod is connected to a flange for connecting the first handling robot (1.1). The second gripper frame includes a third rod and a plurality of fourth rods perpendicular to the third rod. The third rod is connected to a flange for connecting the second handling robot (1.2). Each second rod and fourth rod is connected to a gripper and a camera, and the gripper is connected by a screw structure.

3. The vehicle side panel gripping system according to claim 2, characterized in that, The plurality of second rods are adjustablely connected to the first rod via a ball screw structure. The ball screw is arranged along the extension direction of the first rod, so that the plurality of second rods can remain parallel to each other and their spacing is adjustable.

4. The vehicle side panel gripping system according to claim 2, characterized in that, The plurality of fourth links are adjustablely connected to the third link via a ball screw structure. The ball screw is arranged along the extension direction of the third link, so that the plurality of fourth links can remain parallel to each other and their spacing is adjustable.

5. The vehicle side panel gripping system according to claim 1 or 4, characterized in that, The binocular camera group consists of at least two sets of high-resolution binocular cameras arranged symmetrically, with the baseline distance between each set of binocular cameras set to 300mm. The control system uses a deep learning-based visual guidance algorithm to automatically identify at least 6 key feature points on the board, calculate the displacement deviation of the board relative to the preset reference position along the X, Y, and Z axes and the rotational deviation around the X, Y, and Z axes, and complete the deviation calculation for the six degrees of freedom.

6. The vehicle side panel gripping system according to claim 5, characterized in that, The visual guidance algorithm has a single deviation calculation time of no more than 20ms and an overall positioning accuracy of ±0.25mm.

7. The vehicle side panel gripping system according to claim 1, characterized in that, The control system communicates with the robot via PROFINET industrial Ethernet, with a communication rate of 100Mbps and a signal transmission delay of no more than 5ms.

8. A high-precision, high-load-bearing method for gripping the side of a vehicle body, characterized in that, The method, performed using the system described in any one of claims 1 to 7, comprises the following steps: a. Visual scanning and data processing stage: Drive the first and second handling robots (1.1, 1.2) to move collaboratively to a preset visual scanning range of 500mm to 1500mm away from the side assembly; activate the binocular camera group to take three consecutive pictures of the side assembly according to the Z-shaped scanning trajectory, and take the average of the three measurement results to obtain the final deviation data; b. Deviation Calculation and First Trajectory Correction Stage: Based on the final deviation data, the control system calculates the displacement and rotation deviation of the side assembly in six degrees of freedom, and transmits the deviation data to the control systems of the first and second handling robots (1.1, 1.2) via PROFINET communication to automatically generate and correct their motion trajectories; c. Precise gripping and torque balancing stage: The first and second handling robots (1.1, 1.2) strictly follow the corrected motion trajectory and work together to drive the overall cooperative gripper to precisely fit the preset gripping point of the side assembly. During the gripping process, the load distribution system adjusts the output torque of the two robots in real time to ensure uniform gripping force. d. Dynamic monitoring and secondary trajectory correction stage: During the process of transporting the side assembly from the panel transfer trolley (3) to the assembly fixture (4), the binocular camera group monitors the posture of the side assembly in real time. Once the deviation caused by the posture change exceeds the preset threshold, it immediately feeds back and triggers the first and second handling robots (1.1, 1.2) to perform secondary trajectory correction. e. Precision placement stage: The first and second handling robots (1.1, 1.2) place the side assembly smoothly at the designated position of the assembly fixture (4) according to the final corrected trajectory.

9. The high-precision, high-load vehicle side panel gripping method according to claim 8, characterized in that, In step b, the deviation calculation of the six degrees of freedom is completed by a deep learning-based visual guidance algorithm. This algorithm automatically identifies at least six key feature points on the side assembly and eliminates interference from surface stains or slight deformations in order to calculate the three-dimensional coordinates of each feature point.

10. The high-precision, high-load vehicle side panel gripping method according to claim 8, characterized in that, In step d, the preset threshold for triggering secondary trajectory correction is ≥0.5mm.

Images