An unmanned aerial vehicle mechanical arm adaptive grabbing method and system suitable for automobile parts

Abstract

The application discloses an unmanned aerial vehicle mechanical arm adaptive grabbing method and system suitable for automobile parts, belongs to the technical field of mechanical arm control, and comprises the steps of obtaining part geometric information and initial pose, online identification of local stiffness field, multi-physical quantity coupling pre-tightening force model construction and optimal clamping force calculation, time sequence step-by-step wrapping grabbing, closed-loop monitoring and adjustment of the grabbing process, and adaptive release. The application constructs a local stiffness field of the part by means of clamping exploration and stiffness online solution, and establishes a multi-physical quantity coupling pre-tightening force model in a multi-physical quantity coupling mode, thereby solving the problems that the existing unmanned aerial vehicle mechanical arm grabbing cannot identify the local stiffness distribution and the force control strategy is blind and single, realizing the matching of clamping forces of different regions of automobile special-shaped stiffness parts, and avoiding the problems of crushing of thin-walled regions and slipping of high-rigidity regions.

Description

Technical Field

[0001] This invention belongs to the field of robotic arm control technology, specifically relating to an adaptive grasping method and system for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts. Background Technology

[0002] In the current automotive manufacturing sector, the production, warehousing, and assembly of irregularly shaped rigid parts are continuously increasing. These parts are characterized by irregular curved surfaces, lack of positioning features, highly reflective metal surfaces, and the coexistence of thin-walled and high-rigidity areas. Traditional manual and ground-based fixed robotic arms are inefficient in grasping these parts and are difficult to adapt to the dynamic flow of production lines. While drone robotic arms, as a new type of aerial grasping equipment, are gradually being applied to automotive parts grasping operations, existing technological solutions have significant shortcomings and cannot meet the operational requirements of irregularly shaped rigid parts.

[0003] Existing drone robotic arms lack the ability to identify the local stiffness distribution of parts when performing grasping operations. At the same time, the gripping force control method lacks scientific basis and reasonable planning, and the overall force control strategy is blind and simplistic, which cannot provide stable and reliable aerial grasping support for irregularly shaped and stiff automotive parts. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention aims to provide an adaptive grasping method and system for drone robotic arms adapted to automotive parts, solving the problems of existing drone robotic arms' inability to identify local stiffness distribution and their blindly simplistic force control strategies. The specific solution is as follows: In a first aspect, embodiments of this application provide an adaptive grasping method for a drone robotic arm adapted to automotive parts, comprising the following steps: Acquire the 3D contour data, surface curvature, initial pose information and material type of the parts, and estimate their mass range; The robotic arm end effector is controlled to perform clamping tests on key clamping areas of the parts, collect the changes in clamping force and micro-displacement of the gripper, calculate the local stiffness of the contact point online, scan along the contour of the parts and repeat the clamping tests and calculations to construct the local stiffness field of the parts, and mark the high stiffness area, the secondary stiffness area and the low stiffness thin-walled area. Based on the local stiffness field, surface curvature, mass range, and attitude disturbance data collected by the UAV inertial measurement unit, the surface curvature coefficient, mass offset coefficient, and base disturbance coefficient are calculated. A multi-physical quantity coupled preload model is constructed to calculate the optimal clamping force in the high stiffness region, the secondary stiffness region, and the low stiffness thin-walled region. Based on the local stiffness field distribution and the optimal clamping force in the high stiffness region, the secondary stiffness region and the low stiffness thin-walled region, the main positioning clamping, auxiliary constraint clamping and flexible wrapping clamping are completed step by step in sequence to achieve adaptive wrapping grasping. During flight, the deviation between the actual clamping force at each clamping point and the optimal clamping force in the high-stiffness region, the secondary stiffness region, and the low-stiffness thin-walled region, as well as the change in local stiffness at each contact point, are continuously monitored. The clamping force is adjusted based on the force deviation and the change in stiffness, and the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient. Upon reaching the target placement point, the components are released step by step in the reverse sequence of grasping, and the robotic arm posture and release speed are adjusted to complete adaptive grasping.

[0005] Furthermore, the three-dimensional contour data, surface curvature, initial pose information, and material type of the components are acquired to estimate their mass range, specifically including: The UAV is equipped with a structured light camera and a lidar to perform three-dimensional scanning of the target parts, extract the parts' outlines and reconstruct the geometric models, and obtain three-dimensional outline data and surface curvature. By visually identifying the material type of the parts and combining it with three-dimensional contour data, the volume of the parts is estimated, and then the mass range of the parts is estimated. The initial pose information of the components, including spatial coordinates and attitude angles, is determined by LiDAR point cloud registration.

[0006] Furthermore, the end effector of the robotic arm is controlled to perform gripping probes on the key gripping areas of the components, collecting data on changes in gripping force and micro-displacement of the gripper, and calculating the local stiffness of the contact points online. Specifically, this includes: The robotic arm moves to the preset testing area of ​​the component and controls the end gripper to perform clamping and testing with preset amplitude and preset force. Real-time acquisition of changes in clamping force and small displacement of the gripper during the clamping process; The local stiffness at the contact point is calculated online based on Hooke's Law. The calculation formula is as follows: ; in, For the local stiffness at the contact point, This represents the change in clamping force. This refers to the minute displacement change of the gripper.

[0007] Furthermore, by scanning along the component contour and repeatedly performing clamping trials and calculations, a local stiffness field of the component is constructed, and high-stiffness regions, secondary-stiffness regions, and low-stiffness thin-walled regions are marked, specifically including: The end effector of the robotic arm rapidly scans multiple key gripping areas evenly distributed along the contour of the component. Clamping tests and local stiffness calculations were repeatedly performed in each key clamping area to obtain local stiffness data for each contact point. Summarize the local stiffness data at each contact point to construct the local stiffness field of the component; Based on a preset stiffness threshold, the local stiffness field of the component is divided into a high-stiffness region, a secondary stiffness region, and a low-stiffness thin-walled region.

[0008] Furthermore, the optimal clamping force and base disturbance coefficient are calculated for the high-stiffness region, the secondary-stiffness region, and the low-stiffness thin-walled region, specifically including: The local stiffness of each contact point is obtained based on the local stiffness field, the surface curvature coefficient is calculated based on the surface curvature, the mass offset coefficient is calculated based on the mass range of the components, and the base disturbance coefficient is calculated based on the attitude disturbance data collected by the UAV inertial measurement unit. A multi-physical quantity coupled preload model is constructed by using local stiffness, surface curvature coefficient, mass offset coefficient and base disturbance coefficient as input variables. Substituting local stiffness, surface curvature coefficient, mass offset coefficient, and base disturbance coefficient into the multi-physical quantity coupled preload model, the optimal clamping force is calculated for the high stiffness region, the secondary stiffness region, and the low stiffness thin-walled region, respectively. The optimal clamping force in the high stiffness region is greater than that in the secondary stiffness region, and the optimal clamping force in the secondary stiffness region is greater than that in the low stiffness thin-walled region.

[0009] Furthermore, based on the local stiffness field distribution and the optimal clamping force in each stiffness region, the main positioning clamping, auxiliary constraint clamping, and flexible wrapping clamping are completed step by step in a time sequence to achieve adaptive wrapping grasping, specifically including: Control the robotic arm to move to the high-rigidity area, close the gripper according to the optimal gripping force in the high-rigidity area to complete the main positioning gripping, and establish the gripping reference; Move to the secondary stiffness region and close the gripper to form an auxiliary constraint according to the optimal clamping force of the secondary stiffness region; Adjust the gripper to fit the low-stiffness thin-walled area, and close the gripper with the optimal clamping force of the low-stiffness thin-walled area to achieve flexible wrapping. During the grasping process, the attitude of the drone and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient to compensate for the shaking caused by the base floating.

[0010] Furthermore, the clamping force is adjusted based on the force deviation and stiffness change, and the UAV attitude and robotic arm joint angles are adjusted according to the base disturbance coefficient, specifically including: Real-time monitoring of the actual clamping force at each clamping point, and calculation of the force value deviation from the optimal clamping force in the high-stiffness region, the secondary-stiffness region, and the low-stiffness thin-walled region; Real-time monitoring of changes in local stiffness at each contact point; When the force deviation or stiffness change exceeds the preset threshold, the clamping force is adjusted based on the force deviation and stiffness change, and the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient.

[0011] Furthermore, upon reaching the target placement point, the components are released step-by-step in the reverse sequence of grasping, and the robotic arm posture and release speed are adjusted to complete adaptive grasping. Specifically, this includes: First, release the clamping state of the low-stiffness thin-walled area, and maintain stable support at the remaining points; Release the clamping state of the secondary stiffness region and retain the single-point positioning of the high stiffness region; Finally, the high-rigidity area clamping is released, completing the overall release; During the release process, the robot arm's posture and position are monitored in real time via vision. When deviations occur, the robot arm's posture and release speed are dynamically adjusted.

[0012] Secondly, embodiments of this application provide an adaptive grasping system for a drone robotic arm adapted to automotive parts, comprising: Information acquisition module: used to acquire the three-dimensional contour data, surface curvature, initial pose information and material type of the parts, and estimate their mass range; Stiffness identification module: used to control the end effector of the robotic arm to perform clamping and probing of the key clamping areas of the parts, collect the changes in clamping force and the changes in micro-displacement of the gripper, calculate the local stiffness of the contact point online, scan along the contour of the parts and repeat the clamping and probing and calculation, construct the local stiffness field of the parts, and mark the high stiffness area, the secondary stiffness area and the low stiffness thin-walled area. Force control modeling module: Based on the local stiffness field, surface curvature, mass range and attitude disturbance data collected by the UAV inertial measurement unit, it calculates the surface curvature coefficient, mass offset coefficient and base disturbance coefficient, constructs a multi-physical quantity coupled preload model, and calculates the optimal clamping force in high stiffness region, secondary stiffness region and low stiffness thin-walled region. Step-by-step grasping module: Based on the local stiffness field distribution and the optimal clamping force of high stiffness region, secondary stiffness region and low stiffness thin-walled region, it completes the main positioning clamping, auxiliary constraint clamping and flexible wrapping clamping in a step-by-step manner according to the time sequence to achieve adaptive wrapping grasping; Closed-loop adjustment module: used to continuously monitor the force deviation between the actual clamping force at each clamping point and the optimal clamping force in the high stiffness region, secondary stiffness region and low stiffness thin-walled region during flight, as well as the change in local stiffness at each contact point. The clamping force is adjusted based on the force deviation and stiffness change, and the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient. Adaptive release module: After reaching the target placement point, it releases the parts step by step according to the reverse timing of the grasping, adjusts the robot arm posture and the release speed, and completes adaptive grasping. Beneficial effects

[0013] (1) This invention constructs the local stiffness field of the component by clamping trial and online stiffness calculation, and establishes a multi-physical quantity coupled pre-tightening force model by combining multi-physical quantity coupling method. It solves the problem that the existing UAV robotic arm cannot identify the local stiffness distribution and the force control strategy is blind and single. It realizes the matching of clamping force in different areas of the automotive irregular stiffness component, and avoids the problems of crushing in thin-walled areas and slipping in high-rigidity areas. (2) This invention solves the problem that the floating base disturbance of the UAV is difficult to coordinate and adapt with the grasping action and the conventional clamping method is not stable by performing the sequential step-by-step wrapping and grasping according to the stiffness field distribution and simultaneously compensating for the disturbance of the UAV base. It achieves stable and controllable grasping posture and reliable fixation of parts in the air, and improves the grasping success rate. (3) This invention solves the practical problems of easy fluctuation in clamping state and insufficient placement accuracy of parts during flight by monitoring the changes in clamping force and stiffness in the entire grasping process in a closed loop and performing adaptive adjustment, combined with the step-by-step release of parts in reverse timing. It realizes the automated and stable operation of the entire grasping, handling and placement process, effectively reduces the loss rate of parts and improves the assembly adaptability. Attached Figure Description

[0014] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0015] Figure 1 A flowchart illustrating an adaptive grasping method for a drone robotic arm adapted to automotive parts. Figure 2 This is a structural diagram of an adaptive grasping system for a drone robotic arm adapted to automotive parts. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0017] First, it should be noted that after analyzing the practical application of existing drone robotic arm grasping technology, it was found that traditional drone robotic arm grasping systems generally suffer from problems such as lack of stiffness perception, simplistic force control logic, unreasonable grasping strategies, and insufficient adaptation to floating bases. Existing technologies rely solely on visual recognition of component shape and pose, or on force sensors to monitor total clamping force, failing to identify the local stiffness distribution of components, leading to blind force control strategies. Existing force control often uses fixed force values ​​or empirical thresholds, without combining multiple factors such as component stiffness, surface curvature, and drone floating disturbances in the air, easily resulting in problems such as crushing of thin-walled areas and slippage in high-rigidity areas. Existing grasping strategies use two-point parallel clamping or fixed-point clamping, without optimizing the clamping points and timing according to stiffness distribution, failing to achieve stable and non-destructive grasping. As a floating base, the drone experiences disturbances during flight, and existing technologies do not couple and compensate for base disturbances with clamping force and stiffness perception, leading to wobbling, loosening, or even grasping failure during the grasping process. These combined issues result in a severe lack of stability and reliability for drone robotic arms in scenarios involving the grasping of irregularly shaped and rigid automotive parts, making it difficult to meet the aerial grasping requirements of dynamic automotive production lines.

[0018] See attached document Figure 1 This application provides an adaptive grasping method for automotive parts using a drone-mounted robotic arm. This method uses a drone-mounted robotic arm as the execution unit and achieves stable and non-destructive aerial grasping of irregularly shaped automotive parts through online stiffness field recognition, multi-physical quantity coupled force control, and time-series step-by-step wrapping grasping. The specific implementation of each step is described in detail below with reference to the embodiments.

[0019] I. Obtaining Geometric Information and Initial Pose of Components This step involves performing 3D scanning and geometric reconstruction of the target automotive irregular stiffness component to obtain the component's 3D contour data, surface curvature, initial pose information, and material type, and to estimate its mass range, providing basic data for subsequent stiffness identification and force control calculations.

[0020] It should be noted that irregularly shaped automotive components with high rigidity include irregularly shaped forgings, wheel hubs, cast aluminum parts, and stamped parts. These components are characterized by irregular curved surfaces, lack of positioning features, no obvious texture, and high reflectivity of the metal surface, localized thin walls, and localized high rigidity. The stiffness difference between different areas of the same part can be more than 10 times, with the wheel hub flange area exhibiting high rigidity and the spoke area exhibiting low rigidity. Therefore, this step requires the use of multi-sensor fusion scanning and intelligent recognition algorithms to obtain the geometric and physical property information of the components.

[0021] A drone equipped with a structured light camera and a lidar system performs 3D scanning of target components, extracting their contours and reconstructing their geometric models to obtain 3D contour data and surface curvature. The structured light camera projects structured light stripes onto the component surface, capturing images of the deformed stripes and calculating the surface's 3D coordinates using triangulation. The lidar emits laser pulses and receives reflected signals to measure distances to points on the component's surface. After fusion and registration of the two sensor data, high-density point cloud data is generated, identifying the boundary between the component and the background, extracting the component contours, and reconstructing a complete geometric model, thus obtaining 3D contour data and surface curvature.

[0022] By visually identifying the material type of components and combining it with 3D contour data to estimate the volume of the components, the mass range of the components can be estimated. The visual identification is based on the surface texture, reflectivity, and color characteristics of the components, and uses a pre-trained material classification model to determine the material type, including aluminum alloy, cast iron, and plastic. The volume of the components is calculated by combining it with 3D contour data, and then the mass range of the components is estimated based on the material density.

[0023] The initial pose information of the component, including spatial coordinates and attitude angles, is determined by registration with LiDAR point clouds. The current scanned point cloud is registered with the preset component template point cloud, and the rotation matrix and translation vector are solved to obtain the spatial coordinates and attitude angles of the component in the UAV coordinate system.

[0024] The 3D contour data, surface curvature, initial pose information, and material type are used as input data for subsequent stiffness identification and force control calculations, and are synchronously stored in the cache area of ​​the UAV's onboard computing unit for real-time retrieval in subsequent steps.

[0025] II. Online Identification of Local Stiffness Fields This step involves conducting clamping tests on the key clamping areas of the component, collecting data on changes in clamping force and small displacements of the gripper, calculating the local stiffness of the contact point online, scanning along the component's contour and repeating the clamping tests and calculations to construct the component's local stiffness field, and marking high-stiffness areas, secondary-stiffness areas, and low-stiffness thin-walled areas, providing a basis for subsequent force control calculations and gripping strategy formulation.

[0026] It should be noted that existing drone robotic arms rely solely on visual recognition of component shape and pose, or on force sensors to monitor total gripping force, failing to identify the local stiffness distribution of components. This leads to blind force control strategies and an inability to adapt to the stiffness heterogeneity of the same component. Therefore, this step adopts a pre-grabbing mode of small-scale trial-stiffness calculation-stiffness field construction, achieving for the first time online identification of local stiffness of irregularly shaped automotive components by a drone robotic arm.

[0027] The robotic arm moves to the pre-set testing area of ​​the component, and the end effector gripper is controlled to perform a clamping test with a preset amplitude and preset force. The preset amplitude ranges from 0.1mm to 0.5mm, and the preset force ranges from 5N to 10N. This testing force is much smaller than the actual gripping force to ensure that the surface of the component is not damaged during the testing process.

[0028] The gripping force changes and gripper displacement changes during the gripping process are collected in real time by a six-dimensional torque sensor integrated at the end of the drone's robotic arm. The six-dimensional torque sensor is installed between the end of the robotic arm and the gripper and can simultaneously measure three-axis force and three-axis torque. This step mainly extracts the gripping force change along one axis and the corresponding gripper displacement change.

[0029] The local stiffness at the contact point is calculated online based on Hooke's Law, and the calculation formula is as follows: ; In the formula: This refers to the local stiffness at the contact point, expressed in N / mm. This represents the change in clamping force, expressed in nanometers (N). This represents the small displacement change of the gripper, expressed in mm.

[0030] The robotic arm drives the gripper to rapidly scan multiple key clamping areas evenly distributed along the contour of the component. The key clamping areas are preset to be 6 to 8 areas evenly distributed along the contour of the component, covering the main structural feature areas of the component. For example, the flanges at both ends of a connecting rod forging, the middle of the rod body, and the thin-walled transition area are typical key clamping areas.

[0031] Clamping tests and local stiffness calculations were repeatedly performed in each key clamping area to obtain local stiffness data for each contact point. Each key clamping area underwent 1 to 3 repeated tests, and the average value of the multiple calculations was taken as the local stiffness of the contact point in that area to reduce the error of a single measurement.

[0032] The local stiffness data of each contact point are collected to construct the local stiffness field of the component. The local stiffness of the contact points in each key clamping area is mapped onto the three-dimensional contour model of the component according to its spatial location, forming a local stiffness field distribution map covering the surface of the component.

[0033] Based on preset stiffness thresholds, the local stiffness field of the component is divided into high-stiffness regions, medium-stiffness regions, and low-stiffness thin-walled regions. The preset stiffness thresholds are set as follows: high-stiffness regions correspond to local stiffness ≥ 50 N / mm, medium-stiffness regions correspond to local stiffness > 20 N / mm and < 50 N / mm, and low-stiffness thin-walled regions correspond to local stiffness ≤ 20 N / mm. The division results are used as the basis for subsequent force control calculations and grasping strategy formulation, and are simultaneously marked on the 3D contour model of the component.

[0034] III. Construction of Multi-Physical Quantity Coupled Preload Model and Calculation of Optimal Clamping Force This step is based on the local stiffness field, surface curvature, component mass, and attitude disturbance data collected by the UAV inertial measurement unit. It calculates the surface curvature coefficient, mass offset coefficient, and base disturbance coefficient, constructs a multi-physical quantity coupled preload model, and calculates the optimal clamping force in each stiffness region, so as to achieve scientific calculation of clamping force rather than empirical setting.

[0035] It should be noted that existing force control methods mostly use fixed force values ​​or empirical thresholds, without considering the coupling calculation of multiple factors such as component stiffness, surface curvature, and UAV floating disturbances. This can easily lead to problems such as crushing in thin-walled areas and slippage in high-rigidity areas. Therefore, this step integrates four physical quantities—component local stiffness, surface curvature, UAV base disturbance, and component mass—to construct a dynamic adaptive preload model, solving the problem that single force control cannot adapt to stiffness heterogeneity and floating bases.

[0036] The local stiffness of each contact point is obtained based on the local stiffness field, the surface curvature coefficient is calculated based on the surface curvature, the mass offset coefficient is calculated based on the component mass, and the base disturbance coefficient is calculated based on the attitude disturbance data collected by the UAV inertial measurement unit.

[0037] The surface curvature coefficient is calculated from the three-dimensional contour data of the component. The greater the surface curvature, the greater the surface curvature coefficient. The specific calculation formula is as follows: ; In the formula: The curvature coefficient of the surface; The radius of curvature of the surface at the contact point is in mm. The value range is from 1.05 to 1.8.

[0038] The base disturbance coefficient is calculated from the attitude disturbance data collected in real time by the UAV's inertial measurement unit. The larger the disturbance, the larger the base disturbance coefficient. The specific calculation formula is as follows: ; In the formula: The base disturbance coefficient; This represents the change in pitch angle, expressed in rad. This represents the change in roll angle, expressed in rad. The value range is from 1.0 to 1.5.

[0039] The mass bias coefficient is determined based on the component's material and the volume calculated for visual recognition, after obtaining the component's mass. The specific calculation formula is as follows: ; In the formula: This is the mass bias coefficient; The mass of the component is expressed in kg. The value range is 1.0 to 1.3. The basic friction threshold is determined based on the material of the components and the friction coefficient of the gripper surface. The specific calculation formula is as follows: ; In the formula: The basic friction threshold, in N; For the coefficient of friction between the gripper and the surface of the parts, the coefficient is 0.3 to 0.5 for aluminum alloy, 0.4 to 0.6 for cast iron, and 0.5 to 0.7 for plastic. To determine the minimum clamping force, we take 10N. The value range is from 3N to 7N.

[0040] Substituting the coefficients into the multi-physical quantity coupled preload model, the optimal clamping forces in the high-stiffness region, the secondary-stiffness region, and the low-stiffness thin-walled region are calculated respectively. The calculation formula of the multi-physical quantity coupled preload model is as follows: ; In the formula: The optimal clamping force at a certain contact point, expressed in N; The local stiffness at this contact point, in N / mm, is calculated using the online identification step of the local stiffness field. The curvature coefficient of the surface; The base disturbance coefficient; This is the mass bias coefficient; The basic friction threshold.

[0041] The optimal clamping forces are calculated separately for the high-stiffness region, the secondary-stiffness region, and the low-stiffness thin-walled region. The clamping force in the high-stiffness region is greater than that in the secondary-stiffness region, and the clamping force in the secondary-stiffness region is greater than that in the low-stiffness thin-walled region. This ensures that the clamping force in the high-stiffness region is sufficient to prevent slippage, while the clamping force in the low-stiffness thin-walled region is moderate to avoid crushing.

[0042] IV. Time-sequenced step-by-step package capture This step involves completing the main positioning clamping, auxiliary constraint clamping, and flexible wrapping clamping in a sequential manner based on the local stiffness field distribution and the optimal clamping force in each stiffness region, thereby achieving adaptive wrapping grasping. At the same time, the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient to compensate for base floating.

[0043] It should be noted that existing gripping strategies employ two-point parallel clamping or fixed-point clamping, failing to optimize the clamping points and timing based on stiffness distribution, thus failing to achieve stable and non-destructive gripping. Therefore, this step, based on stiffness field distribution, employs a step-by-step gripping sequence of primary positioning – auxiliary constraint – flexible wrapping, combined with base center of gravity compensation, to achieve stable and non-destructive gripping of irregularly shaped stiff components.

[0044] The robotic arm is moved to a high-rigidity region, and the gripper is closed with the optimal gripping force in that region to complete the main positioning and gripping, establishing a gripping reference. The high-rigidity region has sufficient structural strength to withstand large gripping forces without deformation. Prioritizing gripping the high-rigidity region can quickly establish a stable gripping reference, ensuring the reliability of subsequent auxiliary constraints and flexible wrapping.

[0045] The robotic arm is moved to the secondary stiffness region, and the gripper is closed with the optimal clamping force in that region to form an auxiliary constraint. The stiffness of the secondary stiffness region is between that of the high stiffness region and the low stiffness thin-walled region. Based on the established primary positioning reference, an appropriate clamping force is applied to the secondary stiffness region to form an auxiliary constraint, further improving the gripping stability and preventing the parts from rotating or slipping during flight.

[0046] The gripper is adjusted to conform to the low-stiffness thin-walled region, and then closed with the optimal clamping force for that region to achieve flexible wrapping. The low-stiffness thin-walled region has relatively weak structural strength, and excessive clamping force can easily lead to crushing deformation. Therefore, by adjusting the robotic arm's posture to make the gripper flexibly conform to this region and applying the minimum optimal clamping force, flexible wrapping is achieved, avoiding crushing of the thin-walled region.

[0047] During the grasping process, the drone's attitude and the robotic arm's joint angles are adjusted based on the base disturbance coefficient to compensate for the swaying caused by the base's floating. The drone's inertial measurement unit collects attitude disturbance data in real time and calculates the base disturbance coefficient. When the base disturbance coefficient exceeds a preset threshold, the drone's flight control system adjusts the propeller speed to stabilize the fuselage attitude, and the robotic arm controller synchronously adjusts the angles of each joint to maintain the relative position stability of the gripper and components, ensuring a smooth grasping process.

[0048] V. Closed-loop monitoring and adjustment of the grasping process This step involves continuously monitoring the force deviation between the actual clamping force at each clamping point and the optimal clamping force in each stiffness region during flight, as well as the change in local stiffness at each contact point. The clamping force is adjusted based on the force deviation and stiffness change. At the same time, the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient to maintain stable grasping during flight.

[0049] It should be noted that during the flight of a drone carrying components, factors such as base disturbance, airflow changes, and component inertia can all cause fluctuations in clamping force or changes in stiffness. If not adjusted in time, this can easily lead to loosening, damage, or even the risk of falling. Therefore, this step establishes a closed-loop monitoring mechanism for the flight process, which senses and dynamically adjusts in real time to ensure stable grasping throughout the entire flight.

[0050] The actual clamping force at each clamping point is monitored in real time using a six-dimensional torque sensor, and the force deviation from the optimal clamping force in each stiffness region is calculated. The formula for calculating the force deviation is as follows: ; In the formula: Force deviation, in N; F represents the actual clamping force, in N; F represents the optimal clamping force, in N.

[0051] The change in local stiffness at each contact point is monitored in real time using a stiffness sensor. The formula for calculating the change in stiffness is as follows: ; In the formula: Δk is the change in stiffness, in N / mm; The unit for actual monitoring of local stiffness is N / mm; k is the initial calculated local stiffness, in N / mm.

[0052] When the force deviation or stiffness change exceeds a preset threshold, the clamping force is adjusted based on the force deviation and stiffness change. The preset force threshold is set to 5N, and the preset stiffness ratio threshold is set to 10%. The formula for calculating the adjusted clamping force is as follows: ; In the formula: The adjusted clamping force is expressed in N; λ is the force deviation adjustment coefficient, ranging from 0.8 to 1.2. This is the stiffness variation adjustment coefficient, with a value ranging from 0.5 to 0.8.

[0053] The system adjusts the drone's attitude and the robotic arm's joint angles based on the base disturbance coefficient. When a force deviation or stiffness change exceeds a preset threshold, the base compensation mechanism is triggered simultaneously to adjust the drone's attitude and the robotic arm's joint angles according to the current base disturbance coefficient, eliminating the impact of external disturbances on grasping stability.

[0054] VI. Adaptive Release This step involves releasing the components in stages, following the reverse sequence of the grasping process, after reaching the target placement point, adjusting the robotic arm's posture and the release speed, and completing the adaptive grasping.

[0055] It should be noted that if all clamping points are released simultaneously, the components may tip over or collide due to gravitational imbalance or inertia, resulting in placement deviations or surface damage. Therefore, this step employs a phased release strategy, the opposite of gripping, using real-time visual monitoring and dynamic adjustments to ensure stable placement of the components.

[0056] First, release the grippers in the low-stiffness, thin-walled area: This area experiences minimal clamping force during gripping. Releasing this area first gradually releases the flexible constraints on the component, preventing springback deformation due to sudden stress unloading. Next, release the grippers in the secondary stiffness area: After releasing the low-stiffness, thin-walled area becomes the primary support point. Gradually reducing the clamping force in the secondary stiffness area allows the component's center of gravity to smoothly transition to the high-stiffness area. Finally, release the grippers in the high-stiffness area: As the final support reference, releasing the high-stiffness area last ensures the component maintains a stable posture until fully released, avoiding the risk of overturning due to center of gravity shift.

[0057] During the release process, the placement posture and position of the components are monitored in real time using vision. The structured light camera on the drone continuously collects image data of the placement area, and the visual algorithm identifies the contact state between the bottom surface of the component and the placement plane to determine whether there is any tilting, offset, or suspension.

[0058] When deviations occur, the robotic arm posture and release speed are adjusted to ensure that the parts are placed stably. If visual monitoring detects that the parts are tilting or shifting, the robotic arm controller dynamically adjusts the end effector posture to correct the part's orientation, while reducing the release speed so that the parts fall onto the placement plane slowly and in a controllable manner, ensuring that the placement accuracy meets the assembly requirements of the production line.

[0059] In this application, the drone is a multi-rotor drone with an X-shaped hexacopter layout. The fuselage is made of carbon fiber composite material, the total weight is 8 kg, the maximum takeoff weight is 15 kg, and the endurance is 25 minutes. The drone is equipped with an inertial measurement unit attitude sensor, a structured light camera, and a lidar; the robotic arm is a lightweight multi-joint robotic arm with a six-dimensional torque sensor and an array of flexible haptic pads integrated at the end; the gripper is an adaptive flexible gripper that can automatically conform to the curved surfaces of the components.

[0060] See attached document Figure 2 Based on the same inventive concept, this application also provides an adaptive grasping system for drone robotic arms adapted to automotive parts, comprising: Information acquisition module: used to acquire the three-dimensional contour data, surface curvature, initial pose information and material type of the parts, and estimate their mass range; Stiffness identification module: used to control the end effector of the robotic arm to perform clamping and probing of the key clamping areas of the parts, collect the changes in clamping force and the changes in micro-displacement of the gripper, calculate the local stiffness of the contact point online, scan along the contour of the parts and repeat the clamping and probing and calculation, construct the local stiffness field of the parts, and mark the high stiffness area, the secondary stiffness area and the low stiffness thin-walled area. Force control modeling module: Based on the local stiffness field, surface curvature, mass range and attitude disturbance data collected by the UAV inertial measurement unit, it calculates the surface curvature coefficient, mass offset coefficient and base disturbance coefficient, constructs a multi-physical quantity coupled preload model, and calculates the optimal clamping force in high stiffness region, secondary stiffness region and low stiffness thin-walled region. Step-by-step grasping module: Based on the local stiffness field distribution and the optimal clamping force of high stiffness region, secondary stiffness region and low stiffness thin-walled region, it completes the main positioning clamping, auxiliary constraint clamping and flexible wrapping clamping in a step-by-step manner according to the time sequence to achieve adaptive wrapping grasping; Closed-loop adjustment module: used to continuously monitor the force deviation between the actual clamping force at each clamping point and the optimal clamping force in the high stiffness region, secondary stiffness region and low stiffness thin-walled region during flight, as well as the change in local stiffness at each contact point. The clamping force is adjusted based on the force deviation and stiffness change, and the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient. Adaptive release module: After reaching the target placement point, it releases the parts step by step according to the reverse timing of the grasping, adjusts the robot arm posture and the release speed, and completes adaptive grasping.

[0061] Finally, it should be noted that the above experimental examples are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred experimental examples, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts, characterized in that, Includes the following steps: Acquire the 3D contour data, surface curvature, initial pose information and material type of the parts, and estimate their mass range; The robotic arm end effector is controlled to perform clamping tests on key clamping areas of the parts, collect the changes in clamping force and micro-displacement of the gripper, calculate the local stiffness of the contact point online, scan along the contour of the parts and repeat the clamping tests and calculations to construct the local stiffness field of the parts, and mark the high stiffness area, the secondary stiffness area and the low stiffness thin-walled area. Based on the local stiffness field, surface curvature, mass range, and attitude disturbance data collected by the UAV inertial measurement unit, the surface curvature coefficient, mass offset coefficient, and base disturbance coefficient are calculated. A multi-physical quantity coupled preload model is constructed to calculate the optimal clamping force in the high stiffness region, the secondary stiffness region, and the low stiffness thin-walled region. Based on the local stiffness field distribution and the optimal clamping force in the high stiffness region, the secondary stiffness region and the low stiffness thin-walled region, the main positioning clamping, auxiliary constraint clamping and flexible wrapping clamping are completed step by step in sequence to achieve adaptive wrapping grasping. During flight, the deviation between the actual clamping force at each clamping point and the optimal clamping force in the high-stiffness region, the secondary stiffness region, and the low-stiffness thin-walled region, as well as the change in local stiffness at each contact point, are continuously monitored. The clamping force is adjusted based on the force deviation and the change in stiffness, and the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient. Upon reaching the target placement point, the components are released step by step in the reverse sequence of grasping, and the robotic arm posture and release speed are adjusted to complete adaptive grasping.

2. The adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts as described in claim 1, characterized in that, The acquisition of the three-dimensional contour data, surface curvature, initial pose information, and material type of the component, and the estimation of its mass range, specifically includes: The UAV is equipped with a structured light camera and a lidar to perform three-dimensional scanning of the target parts, extract the parts' outlines and reconstruct the geometric models, and obtain three-dimensional outline data and surface curvature. By visually identifying the material type of the parts and combining it with three-dimensional contour data, the volume of the parts is estimated, and then the mass range of the parts is estimated. The initial pose information of the components, including spatial coordinates and attitude angles, is determined by LiDAR point cloud registration.

3. The adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts according to claim 1, characterized in that, The control system uses the end effector of the robotic arm to perform gripping probes on the key gripping areas of the component, collects changes in gripping force and micro-displacement of the gripper, and calculates the local stiffness of the contact point online. Specifically, this includes: The robotic arm moves to the preset testing area of ​​the component and controls the end gripper to perform clamping and testing with preset amplitude and preset force. Real-time acquisition of changes in clamping force and small displacement of the gripper during the clamping process; The local stiffness at the contact point is calculated online based on Hooke's Law. The calculation formula is as follows: ; in, For the local stiffness at the contact point, This represents the change in clamping force. This refers to the minute displacement change of the gripper.

4. The adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts according to claim 1, characterized in that, The process of scanning along the component contour and repeatedly performing clamping probes and calculations to construct the component's local stiffness field, and marking high-stiffness regions, secondary-stiffness regions, and low-stiffness thin-walled regions, specifically includes: The end effector of the robotic arm rapidly scans multiple key gripping areas evenly distributed along the contour of the component. Clamping tests and local stiffness calculations were repeatedly performed in each key clamping area to obtain local stiffness data for each contact point. Summarize the local stiffness data at each contact point to construct the local stiffness field of the component; Based on a preset stiffness threshold, the local stiffness field of the component is divided into a high-stiffness region, a secondary stiffness region, and a low-stiffness thin-walled region.

5. The adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts according to claim 1, characterized in that, The calculation of the optimal clamping force and base disturbance coefficient for the high-stiffness region, the secondary-stiffness region, and the low-stiffness thin-walled region specifically includes: The local stiffness of each contact point is obtained based on the local stiffness field, the surface curvature coefficient is calculated based on the surface curvature, the mass offset coefficient is calculated based on the mass range of the components, and the base disturbance coefficient is calculated based on the attitude disturbance data collected by the UAV inertial measurement unit. A multi-physical quantity coupled preload model is constructed by using local stiffness, surface curvature coefficient, mass offset coefficient and base disturbance coefficient as input variables. Substituting local stiffness, surface curvature coefficient, mass offset coefficient, and base disturbance coefficient into the multi-physical quantity coupled preload model, the optimal clamping force is calculated for the high stiffness region, the secondary stiffness region, and the low stiffness thin-walled region, respectively.

6. The adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts according to claim 5, characterized in that, Based on the local stiffness field distribution and the optimal clamping force in each stiffness region, the main positioning clamping, auxiliary constraint clamping, and flexible wrapping clamping are completed step-by-step in a time sequence to achieve adaptive wrapping grasping. Specifically, this includes: Control the robotic arm to move to the high-rigidity area, close the gripper according to the optimal gripping force in the high-rigidity area to complete the main positioning gripping, and establish the gripping reference; Move to the secondary stiffness region and close the gripper to form an auxiliary constraint according to the optimal clamping force of the secondary stiffness region; Adjust the gripper to fit the low-stiffness thin-walled area, and close the gripper with the optimal clamping force of the low-stiffness thin-walled area to achieve flexible wrapping. During the grasping process, the attitude of the drone and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient to compensate for the shaking caused by the base floating.

7. The adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts according to claim 6, characterized in that, The adjustment of clamping force based on force deviation and stiffness change, and the adjustment of UAV attitude and robotic arm joint angles based on base disturbance coefficient, specifically include: Real-time monitoring of the actual clamping force at each clamping point, and calculation of the force value deviation from the optimal clamping force in the high-stiffness region, the secondary-stiffness region, and the low-stiffness thin-walled region; Real-time monitoring of changes in local stiffness at each contact point; When the force deviation or stiffness change exceeds the preset threshold, the clamping force is adjusted based on the force deviation and stiffness change, and the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient.

8. The adaptive grasping method for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts according to claim 7, characterized in that, Upon reaching the target placement point, the components are released step-by-step in the reverse sequence of grasping, and the robotic arm posture and release speed are adjusted to complete adaptive grasping. Specifically, this includes: First, release the clamping state of the low-stiffness thin-walled area, and maintain stable support at the remaining points; Release the clamping state of the secondary stiffness region and retain the single-point positioning of the high stiffness region; Finally, the high-rigidity area clamping is released, completing the overall release; During the release process, the robot arm's posture and position are monitored in real time via vision. When deviations occur, the robot arm's posture and release speed are dynamically adjusted.

9. An adaptive grasping system for unmanned aerial vehicle (UAV) robotic arms adapted to automotive parts, characterized in that, include: Information acquisition module: used to acquire the three-dimensional contour data, surface curvature, initial pose information and material type of the parts, and estimate their mass range; Stiffness identification module: used to control the end effector of the robotic arm to perform clamping and probing of the key clamping areas of the parts, collect the changes in clamping force and the changes in micro-displacement of the gripper, calculate the local stiffness of the contact point online, scan along the contour of the parts and repeat the clamping and probing and calculation, construct the local stiffness field of the parts, and mark the high stiffness area, the secondary stiffness area and the low stiffness thin-walled area. Force control modeling module: Based on the local stiffness field, surface curvature, mass range and attitude disturbance data collected by the UAV inertial measurement unit, it calculates the surface curvature coefficient, mass offset coefficient and base disturbance coefficient, constructs a multi-physical quantity coupled preload model, and calculates the optimal clamping force in high stiffness region, secondary stiffness region and low stiffness thin-walled region. Step-by-step grasping module: Based on the local stiffness field distribution and the optimal clamping force of high stiffness region, secondary stiffness region and low stiffness thin-walled region, it completes the main positioning clamping, auxiliary constraint clamping and flexible wrapping clamping in a step-by-step manner according to the time sequence to achieve adaptive wrapping grasping; Closed-loop adjustment module: used to continuously monitor the force deviation between the actual clamping force at each clamping point and the optimal clamping force in the high stiffness region, secondary stiffness region and low stiffness thin-walled region during flight, as well as the change in local stiffness at each contact point. The clamping force is adjusted based on the force deviation and stiffness change, and the attitude of the UAV and the joint angle of the robotic arm are adjusted according to the base disturbance coefficient. Adaptive release module: After reaching the target placement point, it releases the parts step by step according to the reverse timing of the grasping, adjusts the robot arm posture and the release speed, and completes adaptive grasping.

Images