Method for grabbing object by robot and related device

Abstract

The invention discloses a method for grabbing an object by a robot, which comprises the following steps: acquiring surface vibration, transient deformation and three-dimensional deformation data of the object through multi-point pulse excitation, and generating stress wave characteristic mapping data; analyzing internal propagation characteristics of the object according to the stress wave characteristic mapping data, identifying a stress transmission channel, determining a material boundary position and a deformation sensitive area, and generating internal structure characteristic data; calculating a deformation risk coefficient and a stability contribution degree according to the internal structure characteristic data, selecting a grabbing position, and generating grabbing strategy data; and according to the stress wave characteristic mapping data and the grabbing strategy data, the inherent frequency and the local deformation characteristic are identified, a force loading sequence is designed, the vibration state and the deformation amount are monitored, and the force loading sequence is dynamically adjusted. According to the method, the deformation behavior of the object can be accurately predicted, the optimal grabbing strategy is determined by analyzing the internal structure characteristics of the object, dynamic force control is achieved based on frequency modulation, and the instability problem in the grabbing process of the deformed object is effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot grasping and force control, and in particular to a method for robot grasping an object and a related device. Background Art

[0002] Robot grasping objects refers to the process in which the robot establishes stable contact with the target object through its end effector and completes operational tasks such as picking up, moving and placing the object in application scenarios such as industrial automation, agricultural automation, warehousing and logistics, medical assistance and home services. This process requires the robot system to comprehensively consider the geometric characteristics, physical properties, environmental constraints and task requirements of the target object, determine the optimal spatial position, orientation and configuration for the robot end effector, and thus generate a stable, efficient and safe grasping strategy. Accurate grasping posture planning is crucial to ensuring the reliability and efficiency of robot operation, and is also the basic technology for achieving effective interaction between robots and the physical world.

[0003] At present, the mainstream robot grasping planning methods are mainly modeled based on the rigid body assumption, such as grasping evaluation based on force closure, geometric shape matching and sampling optimization. However, in practical applications, a large number of non-rigid objects such as plastic packaging, fabrics, food and medical supplies will undergo local or overall deformation when subjected to grasping force. Although these deformations are small, they will significantly affect the stability and accuracy of grasping. When the robot contacts the deformed object according to the grasping posture designed for the rigid body, the preset contact point position, force distribution and friction conditions will deviate from the ideal state, resulting in unstable grasping, slipping or unnecessary damage to the object. Especially in precision operation scenarios, this grasping failure caused by tiny deformation will seriously affect the system performance. The reason is that traditional grasping planning methods cannot predict the deformation behavior of objects during grasping, and lack the ability to dynamically adjust the grasping strategy according to the real-time deformation state. This technical problem needs to be solved urgently. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problems that the existing robot grasping method cannot predict the deformation behavior of the deformed object during the grasping process and lacks the ability to dynamically adjust the grasping strategy according to the real-time deformation state.

[0005] A first aspect of the present invention provides a method for a robot to grasp an object, the method comprising: By performing multi-point pulse excitation on the object, the surface vibration data, surface transient deformation data and three-dimensional deformation data of the object are obtained, and the acquired data are subjected to time-frequency analysis to extract the parameters of stress wave propagation inside the object, and generate stress wave characteristic mapping data including stress propagation velocity distribution, vibration response amplitude and frequency response characteristics; Analyze the propagation characteristics of stress waves inside the object based on the stress wave characteristic mapping data, identify the stress transmission channels, determine the positions of material boundaries and deformation-sensitive regions inside the object, and generate internal structure characteristic data including stiffness distribution, stress transmission channels, and deformation-sensitive region distribution; Calculate the deformation risk coefficient and stability contribution degree of each region on the object surface according to the internal structure characteristic data, select the grasping positions that are closely connected to the stress transmission channels and avoid the deformation-sensitive regions, and generate grasping strategy data including the grasping force loading positions and initial force loading parameters; Identify the natural frequency and local deformation characteristics of the object at the grasping force loading positions according to the stress wave characteristic mapping data and the grasping strategy data, design the initial force loading sequence, monitor the real-time vibration state and local deformation amount of the object at the grasping force loading positions, and dynamically adjust the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data.

[0006] Preferably, the method for obtaining the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object by performing multi-point pulse excitation on the object, performing time-frequency analysis on the obtained data, and extracting the parameters of stress wave propagation inside the object to generate stress wave characteristic mapping data including stress propagation speed distribution, vibration response amplitude, and frequency response characteristics includes: Select multiple pulse excitation points on the object surface according to the spatial coverage principle, and apply a pulse sequence with a frequency range of 5 - 500 Hz and an amplitude of 0.1 - 0.5 N to the pulse excitation points; Obtain the vibration acceleration data of the object surface according to the pulse sequence, perform Fourier transform on the vibration acceleration data to obtain the spectrum data of each excitation point, and calculate the vibration response amplitude according to the spectrum data; Obtain the displacement change data of the object surface according to the pulse sequence, perform wavelet decomposition on the displacement change data to obtain the surface transient deformation data, and analyze the deformation response of the object surface according to the surface transient deformation data; Obtain the three-dimensional point cloud data of the object according to the pulse sequence, calculate the deformation amount of the object surface to obtain the three-dimensional deformation data; Perform time-frequency joint analysis on the vibration response amplitude, surface transient deformation data, and three-dimensional deformation data to extract the velocity parameter, energy attenuation parameter, and phase parameter of stress wave propagation inside the object; Generate the stress propagation speed distribution according to the velocity parameter, calculate the vibration response amplitude according to the energy attenuation parameter and phase parameter, analyze the frequency response characteristics according to the spectrum data of each excitation point, and generate the stress wave characteristic mapping data.

[0007] Preferably, based on the stress wave characteristic mapping data, analyze the propagation characteristics of stress waves inside the object, identify the stress transfer channels, determine the positions of material boundaries and deformation-sensitive regions inside the object, and generate internal structure characteristic data including stiffness distribution, stress transfer channels, and deformation-sensitive region distribution, including: Calculate the mechanical impedance coefficients of each region of the object according to the stress propagation velocity distribution, analyze the material boundary surfaces inside the object based on the mechanical impedance coefficients, calculate the deformation response parameters of the material boundary surfaces in combination with the vibration response amplitude and frequency response characteristics, and determine the material boundary position data; According to the stress propagation velocity distribution and the vibration response amplitude, apply a preloading force for robot grasping to the surface of the object, analyze the propagation path of the stress wave under the preloading force, identify the bifurcation and convergence nodes and the main transfer channels of the stress wave inside the object, and generate stress transfer channel data; According to the stress transfer channel data and the frequency response characteristics, calculate the dynamic stiffness coefficients of each region inside the object, analyze the distribution law of the dynamic stiffness coefficients along the stress transfer channels, and generate stiffness distribution data; Perform correlation analysis on the stiffness distribution data, stress transfer channel data, and material boundary position data, calculate the deformation response distribution of the object in the robot grasping direction, divide the deformation-sensitive regions based on the deformation response distribution, and generate deformation-sensitive region distribution data; According to the stiffness distribution data, stress transfer channel data, and deformation-sensitive region distribution data, calculate the deformation transfer coefficient of the object under the action of robot grasping, and generate internal structure characteristic data.

[0008] Preferably, according to the stress propagation velocity distribution and the vibration response amplitude, apply a preloading force for robot grasping to the surface of the object, analyze the propagation path of the stress wave under the preloading force, identify the bifurcation and convergence nodes and the main transfer channels of the stress wave inside the object, and generate stress transfer channel data, including: Calculate the preloading contact area on the surface of the object according to the stress propagation velocity distribution, generate a preloading depth gradient map for the preloading contact area, and plan the robot preloading motion trajectory in combination with the vibration response amplitude; Apply a preloading force to the surface of the object along the preloading motion trajectory, obtain the contact force feedback data during the preloading process, adjust the action direction and depth of the preloading force according to the contact force feedback data, and generate preloading force distribution data; Analyze the stress propagation wavefront inside the object according to the preloading force distribution data, identify the wavefront bifurcation points and wavefront convergence points, calculate the attenuation rate of the wavefront propagation in combination with the vibration response amplitude, and generate stress wave propagation path data; Perform connectivity analysis on the stress wave propagation path data, select the propagation paths with attenuation rates lower than the preset threshold as the main transmission channels, and generate stress transmission channel data.

[0009] Preferably, according to the internal structure characteristic data, calculate the deformation risk coefficient and stability contribution degree of each area on the object surface, select the grasping positions that are closely connected to the stress transmission channels and avoid the deformation-sensitive areas, and generate grasping strategy data including the grasping force loading positions and initial force loading parameters, including: According to the stiffness distribution data and deformation-sensitive area distribution data in the internal structure characteristic data, calculate the deformation response range of each area on the object surface in the grasping direction, calculate the deformation transfer coefficient of each area in combination with the stress transmission channels, and generate the deformation risk coefficient based on the deformation response range and the deformation transfer coefficient; According to the stress transmission channel data and stiffness distribution data in the internal structure characteristic data, calculate the control ability of each area on the object surface for stress transmission, and generate the stability contribution degree based on the control ability; According to the deformation risk coefficient and the stability contribution degree, screen candidate grasping areas outside the deformation-sensitive area distribution, calculate the connection node density between the candidate grasping areas and the stress transmission channels, and select the areas where the connection node density meets the grasping stability requirements as the grasping force loading positions; According to the stiffness distribution of the grasping force loading positions, calculate the acting direction and force loading amplitude of the grasping force, and generate the initial force loading parameters; Generate grasping strategy data according to the grasping force loading positions and the initial force loading parameters.

[0010] Preferably, according to the stress wave feature mapping data and the grasping strategy data, identify the natural frequency and local deformation characteristics of the object at the grasping force loading positions, design the initial force loading sequence, monitor the real-time vibration state and local deformation amount of the object at the grasping force loading positions, and dynamically adjust the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data, including: According to the frequency response characteristics in the stress wave feature mapping data and the initial force loading parameters in the grasping strategy data, analyze the transient response characteristics of the object at the grasping force loading positions, and calculate the natural frequency distribution and deformation sensitivity coefficient of the grasping force loading positions; Design the force loading reference trajectory based on the natural frequency distribution and the deformation sensitivity coefficient, calculate the acceleration compensation coefficient during the force loading process, generate the anti-inertia compensation force sequence according to the acceleration compensation coefficient, and superimpose the anti-inertia compensation force sequence on the initial force loading parameters to generate the initial force loading sequence; According to the initial force loading sequence, vibration response data and local deformation data under the action of force loading are obtained, the phase offset of the vibration response data and the strain gradient of the deformation data are calculated, and real-time monitoring data is generated; According to the real-time monitoring data and the deformation sensitivity coefficient, deformation compensation parameters at each grasping force loading position are calculated, the frequency composition of the force loading sequence is adjusted in intervals, and a dynamic force control sequence and a force compensation coefficient are generated; According to the dynamic force control sequence and the force compensation coefficient, the force distribution ratio at each grasping force loading position is calculated, and dynamic compensation is performed on the force distribution parameters.

[0011] Preferably, designing a force loading reference trajectory based on the natural frequency distribution and the deformation sensitivity coefficient, calculating an acceleration compensation coefficient during the force loading process, generating an anti-inertia compensation force sequence according to the acceleration compensation coefficient, and superimposing the anti-inertia compensation force sequence on the initial force loading parameters to generate an initial force loading sequence includes: Determining the motion constraint space of the robot at each grasping force loading position according to the natural frequency distribution, combining the deformation sensitivity coefficient to calculate the pose change range during the loading process, and generating force loading reference trajectory data; Calculating the joint acceleration distribution of the robot at key path points according to the force loading reference trajectory data, performing piecewise smoothing processing on the acceleration distribution, and calculating the inertia compensation parameters at each path point based on the smoothed acceleration distribution to generate an acceleration compensation coefficient; Calculating the magnitude and direction of the inertial force at each grasping force loading position according to the acceleration compensation coefficient, generating anti-inertia compensation force data, and performing phase synchronization processing on the anti-inertia compensation force data to generate an anti-inertia compensation force sequence; Superimposing the anti-inertia compensation force sequence on the initial force loading parameters according to the phase superposition rule to generate an initial force loading sequence.

[0012] The second aspect of the present invention provides a device for a robot to grasp an object. The device for a robot to grasp an object includes: A feature mapping module, configured to obtain the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object by performing multi-point pulse excitation on the object, perform time-frequency analysis on the obtained data, extract the parameters of the stress wave propagating inside the object, and generate stress wave feature mapping data including the stress propagation speed distribution, vibration response amplitude, and frequency response characteristics; A structure analysis module, configured to analyze the propagation characteristics of the stress wave inside the object according to the stress wave feature mapping data, identify the stress transmission channels, determine the material boundary positions and deformation sensitive regions inside the object, and generate internal structure characteristic data including the stiffness distribution, stress transmission channels, and deformation sensitive region distribution; A grasping planning module is used to calculate the deformation risk coefficient and stability contribution of each area on the surface of the object according to the internal structure characteristic data, select a grasping position that is closely connected to the stress transfer channel and avoids deformation sensitive areas, and generate grasping strategy data including grasping force loading position and initial force loading parameters; The force control execution module is used to identify the natural frequency and local deformation characteristics of the object at the grasping force loading position according to the stress wave characteristic mapping data and the grasping strategy data, design an initial force loading sequence, monitor the real-time vibration state and local deformation of the object at the grasping force loading position, and dynamically adjust the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data.

[0013] A third aspect of the present invention provides a device for a robot to grasp objects, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the device for a robot to grasp objects executes the steps of the above-mentioned method for a robot to grasp objects.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the above-mentioned method of robot grasping an object.

[0015] The difficulty of robot grasping non-rigid objects is that the objects will undergo partial or overall deformation during the grasping process, and these deformations cause the preset contact point position, force distribution and friction conditions to deviate from the ideal state. The present invention establishes a complete deformation prediction and control scheme by deeply analyzing the dynamic response characteristics and internal structure of the object.

[0016] Specifically, by applying multi-point pulse excitation to the object and obtaining its surface vibration, transient deformation and three-dimensional deformation data, the dynamic response characteristics of the object under the action of external force can be fully captured. Time-frequency analysis is performed on these data to extract the parameters of stress wave propagation inside the object and generate stress wave feature mapping data, which directly reflects the deformation response characteristics of the object in different regions and at different frequencies. Based on the stress wave feature mapping data, the system can analyze the propagation law of stress waves inside the object, identify stress transfer channels, and determine the material boundary position and deformation sensitive areas. This analysis method breaks through the limitations of traditional reliance on surface features, and accurately predicts the deformation behavior of the object under force through internal structural characteristic data.

[0017] In the grasping planning stage, the system combines internal structural characteristics such as stiffness distribution, stress transmission channels, and deformation-sensitive area distribution to calculate the deformation risk coefficient and stability contribution degree of each area on the object surface. This evaluation method considers the overall structural characteristics of the object, selects positions that are closely connected to the stress transmission channels and avoid deformation-sensitive areas as grasping points, ensuring that the grasping force can be transmitted through the main support structure of the object and reducing the deformation risk.

[0018] In the force control execution stage, the system identifies the natural frequency and local deformation characteristics of the object at the grasping force loading position based on the stress wave characteristic mapping data and grasping strategy data, and designs an initial force loading sequence. By real-time monitoring the vibration state and local deformation amount of the object at the grasping force loading position, the system can timely detect potential unstable factors. According to the monitoring data, the frequency composition and force distribution parameters of the force loading sequence are dynamically adjusted to suppress the generation and development of unstable vibrations. This dynamic force control strategy based on real-time feedback enables the system to respond to the changes in the deformation state during the grasping process at any time.

[0019] This grasping method that combines the internal structural characteristics of the object fundamentally solves the instability problem during the grasping process of deformable objects by accurately predicting the deformation behavior, reasonably planning the grasping strategy, and dynamically adjusting the control parameters. Compared with traditional methods, this solution can not only predict the deformation behavior of the object but also dynamically adjust the grasping strategy according to the real-time monitoring data, significantly improving the reliability and accuracy of grasping deformable objects. Brief Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on the structures shown in these drawings.

[0021] Figure 1 It is a schematic diagram of an embodiment of the method for a robot to grasp an object in an embodiment of the present invention; Figure 2 It is a schematic diagram of an embodiment of the device for a robot to grasp an object in an embodiment of the present invention; Figure 3 It is a schematic diagram of an embodiment of the equipment for a robot to grasp an object in an embodiment of the present invention.

[0022] The realization, functional characteristics, and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the drawings. Detailed Embodiments

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] It should be noted that if there are directional indications (such as up, down, left, right, front, back...) involved in the embodiments of the present invention, the directional indications are only used to explain the relative positional relationship and movement conditions between components in a specific posture. If the specific posture changes, the directional indications will also change accordingly.

[0025] In addition, the descriptions involving "first", "second", etc. in the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In addition, "and / or" throughout the text includes three scenarios. Taking A and / or B as an example, it includes the technical solution of A, the technical solution of B, and the technical solution that both A and B are satisfied at the same time. In addition, the technical solutions between various embodiments can be combined with each other, and it must be based on the fact that those of ordinary skill in the art can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0026] An embodiment of the present application provides a method for a robot to grasp an object. Figure 1 It is a flowchart of a method for a robot to grasp an object provided by an embodiment of the present application. In this embodiment, the method includes: Please refer to Figure 1 , by performing multi-point pulse excitation on the object, obtaining the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object, performing time-frequency analysis on the obtained data, extracting the parameters of the stress wave propagating inside the object, and generating stress wave feature mapping data including stress propagation speed distribution, vibration response amplitude, and frequency response characteristics; In an embodiment of the present invention, the step of performing multi-point pulse excitation on the object, obtaining the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object, performing time-frequency analysis on the obtained data, extracting the parameters of the stress wave propagating inside the object, and generating stress wave feature mapping data including stress propagation speed distribution, vibration response amplitude, and frequency response characteristics includes: Select multiple pulse excitation points on the object surface according to the spatial coverage principle, and apply a pulse sequence with a frequency range of 5 - 500 Hz and an amplitude of 0.1 - 0.5 N to the pulse excitation points; Obtain the vibration acceleration data of the object surface according to the pulse sequence, perform Fourier transform on the vibration acceleration data to obtain the spectral data of each excitation point, and calculate the vibration response amplitude according to the spectral data; Obtain the displacement change data of the object surface according to the pulse sequence, perform wavelet decomposition on the displacement change data to obtain the surface transient deformation data, and analyze the deformation response of the object surface according to the surface transient deformation data; Obtain the three-dimensional point cloud data of the object according to the pulse sequence, calculate the deformation amount of the object surface to obtain the three-dimensional deformation data; Perform time-frequency joint analysis on the vibration response amplitude, surface transient deformation data, and three-dimensional deformation data, and extract the velocity parameter, energy attenuation parameter, and phase parameter of the stress wave propagating inside the object; Generate the stress propagation velocity distribution according to the velocity parameter, calculate the vibration response amplitude according to the energy attenuation parameter and phase parameter, analyze the frequency response characteristics according to the spectral data of each excitation point, and generate the stress wave characteristic mapping data.

[0027] The following is a specific description of the steps involved in the above embodiments: When selecting multiple pulse excitation points on the object surface according to the spatial coverage principle, it is necessary to ensure that the excitation points are evenly distributed and cover the main structural feature areas of the object. In the specific implementation process, first use a three-dimensional scanning device to obtain the geometric features of the object, and then divide the object surface into edge regions, plane regions, and curved surface regions through curvature analysis and topological feature extraction algorithms. For each region, arrange the excitation points according to the principle of spatial uniformity, ensuring that the distance between points is not greater than 1 / 6 of the object size to meet the requirements of the spatial sampling theory. For medical devices with complex shapes, such as syringe assemblies, the excitation point density will be increased at key structural parts such as seams, protrusions, and thin walls. After the excitation points are determined, a pulse sequence with a frequency range of 5 - 500 Hz and an amplitude of 0.1 - 0.5 N is applied using a precision piezoelectric ceramic actuator. This frequency range is selected based on the inherent frequency distribution characteristics of most deformed objects. The low-frequency band (5 - 50 Hz) can effectively excite the overall deformation mode of the object, the middle-frequency band (50 - 200 Hz) corresponds to the local structural response, and the high-frequency band (200 - 500 Hz) can detect the material interface and microstructural characteristics. The amplitude range of 0.1 - 0.5 N ensures that the excitation signal is strong enough to generate a measurable response while avoiding damage to fragile objects, which is particularly important for detecting deformation-sensitive objects such as medical plastic products.

[0028] During the process of obtaining the vibration acceleration data of the object surface, a micro MEMS acceleration sensor array is arranged on the object surface. The sensor sensitivity is not less than 10 mg / √Hz, and the sampling rate is set to 1000 Hz to capture all vibration information in the range of 5 - 500 Hz. The acquired raw acceleration data is first processed by a Butterworth band-pass filter to filter out environmental noise and sensor drift. Subsequently, the fast Fourier transform algorithm is applied to the filtered data to convert the time-domain signal into a frequency-domain representation. To reduce the spectral leakage effect, a Hamming window function is applied to the data before the transformation, and the reliability of the spectral estimation is improved through the 50% overlapping segment averaging technique. For each measurement point, its frequency response function is calculated to analyze the vibration response characteristics at different frequencies. The vibration response amplitude is determined by calculating the root mean square value of the spectral energy, that is, A = √(∑|F(ω)|² / N), where F(ω) is the spectral function and N is the number of frequency points. Taking a flexible plastic package as an example, spectral analysis can clearly identify the differences in resonance characteristics in different regions of the package. For example, the region containing liquid shows obvious low-frequency resonance (15 - 30 Hz), while the edge region of the package shows higher resonance frequencies (80 - 120 Hz). This difference directly reflects the internal structure of the object and the distribution characteristics of the contents.

[0029] The acquisition of surface displacement change data uses a high-speed digital image correlation (DIC) system, which includes a binocular high-speed camera (2000 frames per second) and special image processing software. Before the test, a random black and white speckle pattern is sprayed on the object surface, and the speckle size is controlled within the range of 2 - 5 pixels to ensure the tracking accuracy. The DIC system calculates the displacement vector field of each point on the surface by comparing the speckle images before and after the pulse, achieving a displacement measurement accuracy of 10 μm level. The acquired displacement time series data is subjected to multi-scale analysis using discrete wavelet transform. The Daubechies-4 wavelet basis function is selected for 5-level decomposition to separate the deformation characteristics at different frequency scales. Wavelet decomposition can effectively extract the transient characteristics in the displacement data, especially for the local sudden response in non-linearly deformed objects. Threshold processing is performed on the decomposed wavelet coefficients to remove the influence of noise, and then the surface transient deformation data is reconstructed. For a food container made of multi-layer composite materials, wavelet analysis can clearly distinguish the abnormal transient responses in the regions with poor interlayer adhesion. These regions show obvious high-frequency oscillations and phase lags when subjected to pulse excitation, providing a basis for identifying potential structural weaknesses.

[0030] Three-dimensional point cloud data acquisition uses a structured light three-dimensional scanning system, which consists of a high-precision projector and a binocular camera, with a spatial resolution of up to 0.1 mm. The system projects a phase-shifted sinusoidal fringe pattern onto the object surface, and calculates the three-dimensional coordinates of the surface by analyzing the fringe deformation. To capture the dynamic deformation under pulsed excitation, high-speed projection technology (500 Hz projection frequency) is used to achieve dynamic three-dimensional reconstruction. When processing the original point cloud data, first, the statistical outlier filtering algorithm is applied to remove noise points, and then the iterative closest point algorithm is used to accurately register the point cloud data at different times, eliminating the errors caused by the overall displacement or rotation of the object. The three-dimensional deformation data is calculated by comparing the differences in point cloud coordinates before and after excitation, i.e., ΔP = P1 - P0, where P1 and P0 are the point cloud coordinates after and before excitation, respectively. For the plastic shell of a medical device with a complex shape, three-dimensional point cloud analysis can accurately quantify the deformation distribution in each region, especially the strain concentration phenomenon at the structural transition, such as the deformation amount shown at the connection part under pulsed excitation is 3 - 5 times that of the planar region, and such deformation concentration regions are usually areas that need to be avoided in the grasping plan.

[0031] Time-frequency joint analysis uses a method combining the short-time Fourier transform and the Hilbert-Huang transform to simultaneously examine the evolution characteristics of the signal in the time domain and the frequency domain. The stress wave propagation velocity parameter is obtained by calculating the cross-correlation function of the vibration response signals at different measurement points, finding the time delay τ of the response signals, and then combining the known distance d between the measurement points to obtain the wave velocity v = d / τ. For different regions of the object, the wave velocity parameters are calculated to construct a stress propagation velocity distribution map. The energy attenuation parameter β is obtained by fitting the attenuation curve of the vibration amplitude along the propagation path, and the exponential attenuation model A(d) = is used, where A(d) is the vibration amplitude at a distance d from the propagation start point, is the initial amplitude. The phase parameter φ extracts the instantaneous phase information of the signal through the Hilbert transform and analyzes the phase relationship between each measurement point. These three types of parameters together constitute the characteristic description of the stress wave propagation inside the object. Finally, the thin plate spline interpolation algorithm is applied to extend the parameters of the discrete measurement points to the entire object surface, generating continuous characteristic distribution maps, including the stress propagation velocity distribution map, the energy attenuation distribution map, and the phase distribution map. Combining the spectrum analysis results, a frequency response characteristic map of the object is constructed to form a complete stress wave characteristic mapping data set. Taking a medical soft robot component as an example, time-frequency joint analysis can show the obvious characteristics of a significant reduction in wave velocity (30 - 50% reduction) and an increase in energy attenuation (2 - 3 times increase) at the joint connection, and these regions are usually material transition regions, and their deformation sensitivity needs to be particularly concerned about during the grasping process.

[0032] The generation of stress wave characteristic mapping data is a process of integrating various parameters obtained from the aforementioned analysis into a unified characteristic data set. First, the spatial distribution data of velocity parameters is converted into a stress propagation velocity distribution map, and pseudo-color coding is used to visually display the wave velocity differences in different regions. High wave velocity regions usually correspond to high-stiffness structures inside the object. The vibration response amplitude distribution is calculated by combining the energy attenuation parameter and the phase parameter, reflecting the response intensity of each region of the object to the pulse excitation. The spectral data of each excitation point is used to extract the characteristic frequency and response mode through the principal component analysis method, construct a frequency response characteristic map, and identify the resonance characteristics and dynamic response modes of the object. These characteristic maps are integrated into a multi-level stress wave characteristic mapping data set through a data fusion algorithm, including a time-domain characteristic layer, a frequency-domain characteristic layer, and a spatial distribution characteristic layer. For complex plastic products such as the outer shell of a medical syringe pump, the complete characteristic mapping can clearly show the sudden change in wave velocity (change rate > 50%) and the enhanced energy absorption (attenuation coefficient increases by 2 - 4 times) at the material boundary, revealing the discontinuity of the internal structure and potential deformation-sensitive regions, providing a key basis for subsequent grasping planning. This multi-dimensional characteristic mapping method breaks through the limitations of traditional physical models, directly extracts the structural characteristics from the actual mechanical responses of objects, and can meet the grasping planning requirements of various complex deformed objects.

[0033] Please continue to refer to Figure 1 , analyze the propagation characteristics of stress waves inside the object according to the stress wave characteristic mapping data, identify the stress transmission channels, determine the position of the material boundary inside the object and the deformation-sensitive regions, and generate internal structure characteristic data including stiffness distribution, stress transmission channels, and deformation-sensitive region distribution; In an embodiment of the present invention, the analyzing the propagation characteristics of stress waves inside the object according to the stress wave characteristic mapping data, identifying the stress transmission channels, determining the position of the material boundary inside the object and the deformation-sensitive regions, and generating internal structure characteristic data including stiffness distribution, stress transmission channels, and deformation-sensitive region distribution includes: Calculate the mechanical impedance coefficient of each region of the object according to the stress propagation velocity distribution, analyze the material boundary surface inside the object based on the mechanical impedance coefficient, calculate the deformation response parameters of the material boundary surface by combining the vibration response amplitude and the frequency response characteristics, and determine the material boundary position data; Apply a preloading force for robot grasping to the surface of the object according to the stress propagation velocity distribution and the vibration response amplitude, analyze the propagation path of stress waves under the preloading force, identify the bifurcation and convergence nodes and the main transmission channels of stress waves inside the object, and generate stress transmission channel data; Calculate the dynamic stiffness coefficient of each region inside the object according to the stress transmission channel data and the frequency response characteristics, analyze the distribution law of the dynamic stiffness coefficient along the stress transmission channels, and generate stiffness distribution data; Perform a correlation analysis on the stiffness distribution data, stress transfer channel data, and material boundary position data, calculate the deformation response distribution of the object in the robot grasping direction, divide the deformation-sensitive areas based on the deformation response distribution, and generate deformation-sensitive area distribution data; According to the stiffness distribution data, stress transfer channel data, and deformation-sensitive area distribution data, calculate the deformation transfer coefficient of the object under the action of robot grasping, and generate internal structure characteristic data.

[0034] The following specifically describes the steps involved in the above embodiments: When calculating the mechanical impedance coefficient of each region of the object according to the stress propagation speed distribution, it is first necessary to understand the physical meaning of the mechanical impedance coefficient, that is, the impedance characteristic of the material to the propagation of mechanical waves, which reflects the influence degree of the internal structure of the material on the stress wave transmission. During the calculation process, combine the stress propagation speed distribution data with the object density distribution data to calculate the mechanical impedance coefficient of each region. The calculation formula is Z = ρv, where Z is the mechanical impedance coefficient, ρ is the local density, and v is the stress propagation speed. The density distribution data can be estimated through the volume-mass relationship of the object or obtained using a radiation density measurement device. For objects with complex internal structures, such as the outer shell of an electronic device formed by multi-material injection molding, the calculated mechanical impedance coefficient will exhibit obvious mutation characteristics at the junction of different materials. Perform a gradient analysis on the mechanical impedance coefficient spatial distribution data to identify regions where the gradient change exceeds 30%. These regions usually correspond to the material interfaces. Further combine the vibration response amplitude and frequency response characteristics to calculate the deformation response parameters of the material interface, including the deformation transfer coefficient and the phase difference coefficient. The deformation transfer coefficient represents the ratio of the vibration amplitudes on both sides of the boundary, reflecting the transmission efficiency of the boundary to the vibration energy; the phase difference coefficient represents the difference in vibration phases on both sides of the boundary, reflecting the delay or advance effect of the boundary on the vibration wave. Through the comprehensive analysis of these parameters, determine the precise position, direction, and mechanical coupling characteristics of the material boundary, and form the material boundary position data. For medical soft robots, this step can accurately identify the silicone-metal joint surface and the transition region between silicones with different hardnesses inside them. These boundary regions are often stress concentration and deformation-prone regions during the grasping process, directly affecting the grasping stability.

[0035] During the process of applying the preloading force for robot grasping on the object surface according to the stress propagation speed distribution and the vibration response amplitude, first, determine the preloading contact area on the object surface based on the stress propagation speed distribution, and preferentially select the area with relatively uniform stress propagation speed as the preloading points. Subsequently, generate a preloading depth gradient map in these preloading point areas, where the depth gradient ranges from 0.2 mm to 2 mm and the distribution follows a Gaussian profile to avoid edge stress concentration. Combining the vibration response amplitude data, plan the robot preloading motion trajectory to ensure that the important structural features of the object are not disturbed during the preloading process. The end effector of the robot is equipped with a high-precision force sensor with a resolution of 0.01 N to accurately control the magnitude and direction of the preloading force. During the preloading execution, the robot contacts the object surface at a low speed of 0.5 - 2 mm / s and applies a preloading force of 0.5 - 2 N, while recording the contact force feedback data. According to the contact force feedback data, adjust the action direction and depth of the preloading force in real time to avoid overpressure on the deformation-sensitive areas. The adjustment method follows the principle of adaptive impedance control. When a sudden change in the contact force is detected, the system immediately reduces the applied force and adjusts the contact angle. After completing the preloading, use the sensing system in step one to record the stress wave propagation characteristics in the preloading state and analyze the propagation path of the stress wave inside the object. By comparing the differences in the stress wave propagation characteristics at the same excitation point before and after loading, identify the bifurcation points and convergence points of the stress wave. The bifurcation point refers to the position where the stress wave splits into multiple wavefronts during propagation, usually appearing in the area with reduced stiffness; the convergence point is the position where multiple wavefronts converge, often corresponding to the area with enhanced stiffness. Combine the vibration response amplitude to analyze the attenuation rate of the wavefront propagation, and select the propagation path with an attenuation rate lower than 0.3 dB / cm as the main transmission channel to generate the stress transmission channel data. For the aviation components made of composite materials, this step can reveal the distribution trend of the internal reinforcing fibers and the mechanical transmission characteristics, ensuring that the grasping force can be distributed along these high-strength channels rather than acting on the fragile resin matrix between the fibers.

[0036] When calculating the dynamic stiffness coefficients of each region inside an object based on the stress transfer channel data and the frequency response characteristics, a dynamic testing method combined with frequency response analysis is adopted. The dynamic stiffness coefficient refers to the ability of an object to resist deformation under dynamic loads. Different from static stiffness, it takes into account the inertial effects and damping characteristics of the material. In the calculation process, a standardized sinusoidal excitation force (force amplitude 0.1 - 0.5 N, frequency covering the range of 10 - 300 Hz) is first applied to the key points on the stress transfer channel, and at the same time, the displacement responses of the corresponding points are measured to construct the frequency response function of each test point. By analyzing the amplitude and phase characteristics of the frequency response function, the dynamic stiffness coefficient of each point is extracted, and the distribution law of this coefficient along the stress transfer channel is calculated. For regions where the change in the dynamic stiffness coefficient is significant, the test point density is increased to improve the resolution. Special attention is paid to the change trend of the dynamic stiffness at different frequencies because many deformed objects exhibit different stiffness characteristics at different frequencies. For example, polymeric materials are softer at low frequencies and harder at high frequencies. Based on the measurement results, a spatial interpolation algorithm is used to generate the dynamic stiffness distribution data covering the entire object. For thin-walled tubular objects such as medical catheters, the distribution of the dynamic stiffness coefficient can reveal the non-uniformity of the strength distribution of the catheter wall within the operating frequency range (usually 20 - 100 Hz), especially the stiffness mutation at the joints. This information is crucial for determining safe and reliable grasping positions and can avoid local excessive deformation or damage caused by stiffness mismatch.

[0037] In the process of correlative analysis of the stiffness distribution data, stress transfer path data, and material boundary position data, a three-dimensional data fusion model is first constructed to align these three types of data in the same spatial coordinate system. The correlative analysis uses the multivariate correlation analysis method to calculate the spatial correlation coefficients between the three types of data and identify regions with strong mutual correlations. Based on the fused data, the deformation response distribution of the object under the action of preset grasping forces in different directions (within the range of 1 - 5 N) is simulated and calculated. The deformation response distribution is calculated by the finite element analysis method combined with the measured material parameters, representing the expected deformation amounts of each region of the object under the given grasping force. The calculation process pays particular attention to the deformation responses in the regions where the stress transfer paths cross the material boundaries, which are often high-risk points of discontinuous deformation. According to the calculated deformation response distribution, a deformation sensitivity threshold is set (usually set to 1.5 times the average deformation amount), and the regions with deformation amounts exceeding the threshold are classified as deformation-sensitive regions. For regions with complex shapes, the region growing algorithm is used to expand the boundaries of the sensitive regions to ensure coverage of potential deformation risk regions. The finally generated deformation-sensitive region distribution data includes information such as the spatial positions, ranges, sensitivity levels, and main deformation directions of the sensitive regions. Taking a food product with liquid in a packaging bag as an example, the correlative analysis can accurately identify the deformation-sensitive regions at the junctions of the liquid aggregation regions and the weak points of the packaging material. These regions are extremely prone to the risk of excessive local deformation during the grasping process, leading to packaging rupture, and special avoidance or special force control strategies need to be adopted in the grasping planning.

[0038] When calculating the deformation transfer coefficient of an object under the action of a robotic grasp according to the stiffness distribution data, stress transfer channel data, and deformation-sensitive area distribution data, it is first necessary to understand the physical meaning of the deformation transfer coefficient. The deformation transfer coefficient describes how the deformation of one region in an object affects adjacent regions and reflects the propagation characteristics of deformation within the object. The calculation process uses structural dynamics analysis methods, with the stress transfer channel as the main line, to calculate the strain distribution transmitted along the channel when a unit deformation is applied. In specific implementation, a standardized displacement excitation (usually 0.5 mm) is applied at the selected grasp point on the object surface, the strain distribution along the stress transfer channel is measured, and the strain ratio of each point relative to the excitation point is calculated, which is the deformation transfer coefficient. At the same time, considering the deformation transfer characteristics in different directions, the tensile, shear, and torsional deformation transfer coefficients along the stress transfer channel are calculated respectively. Particularly for the deformation-sensitive area, the deformation coupling relationship with the surrounding structure is analyzed in detail, and the indirect influence of the grasping force acting on the deformation-sensitive area through the stress transfer channel is evaluated. Finally, all the analysis results are integrated to generate internal structure characteristic data including stiffness distribution, stress transfer channel topology, deformation-sensitive area position, and deformation transfer relationship. This data set comprehensively describes the mechanical behavior characteristics of the object under the action of a robotic grasp and provides a direct basis for subsequent grasping planning. For a disposable medical syringe, this step can reveal the deformation transfer relationship between the piston and the barrel. If the grasping force acts along a specific stress transfer channel, even if the grasp point is far from the piston, it may indirectly cause the piston to move through the channel with a high deformation transfer coefficient, resulting in the risk of accidental injection of the liquid medicine. Guided by the internal structure characteristic data, the robot can select a grasping scheme with a low deformation transfer coefficient to avoid such risks.

[0039] In an embodiment of the present invention, when applying a preloading force of a robotic grasp to the object surface according to the stress propagation speed distribution and the vibration response amplitude, analyzing the propagation path of the stress wave under the preloading force, and identifying the bifurcation and convergence nodes and the main transfer channels of the stress wave inside the object to generate stress transfer channel data, including: Calculating the preloading contact area on the object surface according to the stress propagation speed distribution, generating a preloading depth gradient map for the preloading contact area, and planning the preloading motion trajectory of the robot in combination with the vibration response amplitude; Applying a preloading force to the object surface along the preloading motion trajectory, obtaining the contact force feedback data during the preloading process, and adjusting the acting direction and depth of the preloading force according to the contact force feedback data to generate preloading force distribution data; Analyzing the stress propagation wavefront inside the object according to the preloading force distribution data, identifying the wavefront bifurcation points and wavefront convergence points, and calculating the attenuation rate of the wavefront propagation in combination with the vibration response amplitude to generate stress wave propagation path data; Perform a connectivity analysis on the stress wave propagation path data, select the propagation paths with an attenuation rate lower than the preset threshold as the main transmission channels, and generate stress transmission channel data.

[0040] The following is a specific description of the steps involved in the above embodiments: In the process of calculating the preloaded contact area on the object surface according to the stress propagation speed distribution, first perform a threshold segmentation process on the stress propagation speed distribution map, identify the area where the stress propagation speed is higher than 1.2 times the average value as the structurally stable area, and the area where the speed is lower than 0.8 times the average value as the deformation-sensitive area. In the structurally stable area, select the area with an area greater than 20 mm² and relatively flat (curvature change less than 0.1 ) as the candidate preloaded contact area. Score these candidate areas, and the scoring indicators include the distance from the center of the object, the stability of the surface normal vector, and the stress propagation speed gradient in the surrounding area. Select the top 3-5 areas with the highest scores as the preloaded contact areas. For the selected preloaded contact areas, generate a preloading depth gradient map according to their surface shapes and stress propagation characteristics, which specifies the contact depth values at different positions during the robot's preloading. The depth gradient gradually decreases from the center of the contact area to the outside, usually with a maximum depth of 0.5-2 mm at the center and a minimum depth of 0.1-0.5 mm at the edge, forming a smoothly varying pressure distribution. Combine the vibration response amplitude data to optimize the preloading trajectory within the contact area and avoid the points with abnormally high vibration response amplitudes (these points usually correspond to the unstable internal structure areas). The trajectory planning uses the B-spline curve algorithm to ensure smooth and continuous movement of the robot and avoid the impact caused by rapid changes. Taking a medical hose as an example, the stress propagation speed analysis shows that the wave speed in the area with uniform wall thickness is significantly higher than that at the joint, so the system automatically selects the area with uniform wall thickness as the preloaded contact area and generates a depth gradient map that gradually changes from the center to both sides to ensure that the preloading force is mainly distributed in the most stable area of the structure.

[0041] The process of applying a preloading force to the object surface along the preloading motion trajectory is performed by a high-precision force control robot equipped with a torque sensor with a resolution of 0.01 N and a response frequency of not less than 1 kHz. The robot approaches the object according to the preset trajectory, and the initial contact adopts a low-speed mode (below 5 mm / s) to avoid contact impact. After contact is established, the robot gradually increases the contact depth according to the preloading depth gradient map while obtaining real-time contact force feedback data. The contact force feedback data includes three-dimensional force vectors and torque data, with a sampling frequency of 500 Hz, recording the mechanical response characteristics of the object surface during the preloading process. According to the contact force feedback data, the acting direction and depth of the preloading force are adjusted in real time. The adjustment strategy is based on the force control impedance model. When a sudden change in contact force (change rate exceeding 2 N / s) is detected, the system immediately reduces the contact depth and adjusts the contact angle to avoid damaging the object. When the contact force stabilizes within the target range (usually 1 - 3 N) for more than 100 ms, the current force-position data pair is recorded to generate preloading force distribution data. The preloading force distribution data is represented in the form of a three-dimensional force field, including the force magnitude, direction, and corresponding object surface deformation amount at each point in the contact area. Taking flexible food packaging as an example, during the preloading process, the force-feedback data shows that the force-displacement curve in the edge area of the packaging is significantly steeper, while the middle filling area is relatively flat. The system automatically adjusts the preloading depth accordingly, reducing the depth by 30% in the edge area to avoid the risk of excessive deformation.

[0042] When analyzing the stress propagation wavefront inside an object based on the preloading force distribution data, first, under the preloading state, a standardized pulse excitation (frequency 50 - 200 Hz, amplitude 0.2 N) is applied to the object, and the propagation process of the stress wave on the object surface is captured by a high-speed imaging system and a surface vibration sensor array. The stress propagation wavefront refers to the wavefront formed when the stress wave propagates inside the object. The internal wavefront propagation can be indirectly inferred by analyzing the spatio-temporal distribution characteristics of the surface vibration. The specific analysis method uses the time-domain cross-correlation technique to calculate the time delay of the vibration signals between different measurement points, and combines the spatial distribution of the measurement points to reconstruct the propagation path and velocity distribution of the stress wave. During the reconstruction process, special attention is paid to the regions where the propagation direction and velocity of the stress wave change abruptly. These regions usually correspond to the wavefront bifurcation points or wavefront convergence points. The wavefront bifurcation point is the position where the stress wave splits into multiple wavefronts during propagation, usually appearing at the interface where the internal structure of the object changes from dense to loose or from rigid to flexible; the wavefront convergence point is the position where multiple wavefronts converge, commonly found in the regions where the internal structure changes from loose to dense or from flexible to rigid. Combining the vibration response amplitude data, calculate the attenuation rate of the wavefront propagation, that is, the degree of amplitude attenuation of the wavefront during propagation, expressed in decibels per centimeter (dB / cm). According to the calculation results, generate the complete stress wave propagation path data, including the wavefront propagation velocity map, bifurcation point distribution map, convergence point distribution map, and attenuation rate distribution map. For the medical syringe assembly, the preloading analysis can clearly show the stress wave bifurcation phenomenon from the barrel to the needle connection, where the attenuation rate increases significantly (>1.5 dB / cm), indicating that this is a region with significant structural changes and needs to be specially considered in the grasping planning.

[0043] When performing connectivity analysis on the stress wave propagation path data, the graph theory algorithm is used to regard the interior of the object as a stress propagation network, the measurement points as network nodes, and the wavefront propagation paths as connecting edges. First, a weighted directed graph model is constructed. The weight of an edge is determined by the attenuation rate of the stress wave. The lower the attenuation rate, the higher the propagation efficiency, and the greater the corresponding edge weight. The Dijkstra shortest path algorithm is used to analyze the optimal stress transfer path between any two points, that is, the propagation path with the lowest attenuation rate. By calculating the connectivity indicators of the network, such as the average path length, clustering coefficient, and node degree distribution, the overall efficiency and local characteristics of stress transfer inside the object are evaluated. Particular attention is paid to the key nodes in the network. These nodes are located at the intersections of multiple transfer paths and control the global stress distribution. The core step of connectivity analysis is to set a preset threshold to screen the main transfer channels. The preset threshold is the upper limit of the attenuation rate set based on the specific application scenario. Usually, a range of 0.3 - 0.5 dB / cm is selected. This range can ensure that the selected transfer channels have a sufficiently high energy transfer efficiency while excluding high-attenuation regions. The propagation paths with an attenuation rate lower than the preset threshold are selected as the main transfer channels. These channels represent the stress path network with the highest transfer efficiency inside the object and are the preferred channels for the transfer of mechanical loads inside the object. The finally generated stress transfer channel data is represented in the form of a three-dimensional network, including the spatial distribution, attenuation characteristics, and topological connection relationships of the main channels. Taking a food packaging made of multi-layer composite materials as an example, connectivity analysis can identify the orientation of high-strength fibers in the packaging material. The attenuation rate of the main transfer channels formed by these fibers is lower than 0.2 dB / cm, while the attenuation rate in the non-reinforced area is as high as 0.8 dB / cm. The grasping plan should preferably select areas that are well-connected to these main transfer channels to ensure that the grasping force can be effectively distributed and avoid local stress concentration.

[0044] Please continue to refer to Figure 1 According to the internal structure characteristic data, calculate the deformation risk coefficient and stability contribution degree of each area on the surface of the object, select the grasping positions that are closely connected to the stress transfer channels and avoid the deformation-sensitive areas, and generate the grasping strategy data including the grasping force loading positions and initial force loading parameters; In an embodiment of the present invention, the calculating the deformation risk coefficient and stability contribution degree of each area on the surface of the object according to the internal structure characteristic data, selecting the grasping positions that are closely connected to the stress transfer channels and avoiding the deformation-sensitive areas, and generating the grasping strategy data including the grasping force loading positions and initial force loading parameters includes: According to the stiffness distribution data and deformation-sensitive area distribution data of the internal structure characteristic data, calculate the deformation response range of each area on the surface of the object in the grasping direction, combine the stress transfer channels to calculate the deformation transfer coefficient of each area, and generate the deformation risk coefficient based on the deformation response range and the deformation transfer coefficient; Calculate the control ability of each area on the object surface for stress transfer according to the stress transfer channel data and stiffness distribution data of the internal structure characteristic data, and generate a stability contribution degree based on the control ability; According to the deformation risk coefficient and the stability contribution degree, screen candidate grasping areas in the outer periphery of the deformation-sensitive area, calculate the connection node density between the candidate grasping areas and the stress transfer channels, and select the area where the connection node density meets the grasping stability requirements as the grasping force loading position; Calculate the acting direction and force loading amplitude of the grasping force according to the stiffness distribution of the grasping force loading position, and generate initial force loading parameters; Generate grasping strategy data according to the grasping force loading position and the initial force loading parameters.

[0045] The following specifically describes the steps involved in the above embodiments: When calculating the deformation response range of each area on the object surface according to the stiffness distribution data and deformation-sensitive area distribution data of the internal structure characteristic data, first uniformly select test points on the object surface, and the test point density is 3-5 points per square centimeter. Apply a standardized test force (usually 1N) to each test point, and apply it respectively along the main possible grasping directions (such as the x, y, and z coordinate axes directions). Measure the local deformation amount at this time through a high-precision displacement sensor. The deformation response range is defined as the displacement range generated on the object surface under the action of a unit force, reflecting the difference in the ability of the object to resist deformation in different directions. Combining the stress transfer channel data, calculate the deformation transfer coefficient of each area, which represents how local deformation is transmitted to other parts of the object through the internal structure. The specific calculation method is to apply a unit displacement at the test point and measure the response displacement of the key points along the stress transfer channel, and calculate the ratio of the response displacement to the applied displacement. Based on the deformation response range and the deformation transfer coefficient, use the weighted summation method to generate the deformation risk coefficient, and the calculation formula is: risk coefficient = α × maximum deformation response + β × deformation transfer coefficient, where α and β are weight coefficients and are adjusted according to specific application scenarios. Taking a medical injection pump as an example, the deformation response range of the stiffness mutation area at the connection between the infusion tube and the pump body is 2-3 times larger than that of other areas under the same test force, and the deformation transfer coefficient is as high as 0.8, resulting in its deformation risk coefficient being significantly higher than that of other areas, directly affecting the injection accuracy. Therefore, such high-risk areas need to be avoided in the grasping plan.

[0046] When calculating the stress transfer control ability of each region on the object surface based on the stress transfer channel data and stiffness distribution data of the internal structure characteristics, first construct a stress transfer network topology model, regarding the object as a network composed of nodes and edges. Each stress transfer channel is represented as an edge in the network, and the channel intersection points are represented as nodes. Calculate the connectivity of each surface region to the stress transfer network, which is jointly determined by the number and quality (determined by the stress transfer efficiency of the channel) of the stress transfer channels directly connected to the region. Apply the network centrality analysis algorithm to calculate the position importance of each surface region in the entire stress transfer network, including degree centrality (the number of directly connected channels), betweenness centrality (the number of shortest paths passing through this point), and eigenvector centrality (a comprehensive index considering the importance of adjacent points). The control ability refers to the degree of influence of the region on the overall stress distribution when it is used as a grasping point, which is jointly determined by the centrality index of the region in the stress transfer network and the local stiffness of the region. Calculate the stability contribution based on the control ability. The higher the stability contribution, the better the overall stable control that the region can provide as a grasping point. Taking a medical soft robot finger as an example, the analysis shows that although the local stiffness of the region near the joint is low, due to being located at the intersection of multiple stress transfer channels, it has extremely high control ability, and its stability contribution is more than twice that of the general region, which explains why when grasping a soft finger, even applying a slight force at the joint can effectively control the entire component.

[0047] In the process of screening candidate grasping regions around the periphery of the deformation-sensitive region according to the deformation risk coefficient and stability contribution, first exclude the regions where the deformation risk coefficient is higher than the safety threshold (usually set to 1.5 times the average risk coefficient). Among the remaining regions, select the regions with a stability contribution higher than the minimum requirement (usually 60% of the maximum stability contribution) as the preliminary candidate regions. To ensure the safety boundary, set a buffer zone around the deformation-sensitive region, and the width of the buffer zone is set according to the object size ratio, generally 5%-10% of the maximum size of the object. Among the candidate regions that meet the above conditions, calculate the connection node density of each region to the stress transfer channels, that is, the number of nodes directly connected to the main stress transfer channels per unit area. The connection node density reflects the coupling degree of the region with the internal support structure of the object, and the higher the density, the easier it is for the grasping force to be evenly distributed through the internal structure. Select the regions with a connection node density greater than the specified threshold (usually 2-3 nodes per square centimeter) as the final grasping force loading positions. For highly deformation-sensitive objects such as medical catheters, the analysis shows that although the joint region has a relatively high stability contribution, its deformation risk coefficient is also significantly higher than that of the homogeneous region. Therefore, finally, the tube body region 2-3 cm away from the joint is selected as the grasping position, and the connection node density of this region reaches 2.5 nodes per square centimeter, ensuring that the grasping force can be evenly distributed without causing local excessive deformation.

[0048] When calculating the acting direction and force loading amplitude of the grasping force based on the stiffness distribution at the grasping force loading position, first analyze the stiffness tensor characteristics at this position to identify the maximum stiffness direction and the minimum stiffness direction. The stiffness tensor is obtained by applying small displacements in different directions and measuring the corresponding reaction forces, which reflects the object's resistance to deformation in different directions. The acting direction of the grasping force is preferably selected to be consistent with the maximum stiffness direction, so as to make the best use of the object's internal support structure and reduce the risk of deformation. For objects made of anisotropic materials, such as carbon fiber reinforced composites, the stiffness along the fiber direction may be 5-10 times that perpendicular to the fiber direction, and choosing the correct grasping direction is crucial. The force loading amplitude is calculated based on the object's mass, friction coefficient, and safety factor to ensure sufficient grasping stability while avoiding excessive force. For an object with a mass of m, considering the gravitational acceleration g and the safety factor k (usually taken as 1.5-2), the basic grasping force is calculated as F = k×m×g. According to the friction coefficient μ between the object surface and the gripper (usually between 0.3-0.8, depending on the surface material), the actual grasping force Freal = F / μ is adjusted. Taking a soft food package as an example, when applying the grasping force along the texture direction of the packaging material, only 1.2 times the force in the vertical direction is required to achieve the same grasping stability, and the deformation amount is reduced by about 40%, significantly improving the grasping safety and accuracy.

[0049] In the process of generating the grasping strategy data based on the grasping force loading position and the initial force loading parameters, first determine the optimal gripper configuration and contact mode. According to the geometric characteristics and surface conditions of the grasping force loading position, select a suitable end effector (such as parallel jaws, suction cups, or flexible grippers) and contact mode (point contact, line contact, or surface contact). Construct a timing plan for grasping execution, including the approach phase, contact establishment phase, force increase phase, and stable grasping phase. For each phase, specify in detail the robot's motion parameters (speed, acceleration, and attitude change) and force control parameters (target force, stiffness coefficient, and damping coefficient). Especially for deformable objects, in the force increase phase, adopt a progressive force loading curve, and control the force growth rate within the range of 0.5-2 N / s to avoid unstable deformation caused by sudden impacts. Considering the mass distribution and expected motion trajectory of the object, calculate the maximum inertial force that may be encountered during the grasping process and the required anti-inertial compensation. The finally generated grasping strategy data includes the grasping position coordinates, grasping attitude matrix, force loading curve, contact mode parameters, and dynamic adjustment rules. Taking the flexible sample tube in medical testing equipment as an example, the grasping strategy data specifies a two-point grasping mode, with the two grasping points located on both sides of the stable area of the tube body, the initial contact force is 0.5 N, and it increases to 1.5 N at a rate of 0.8 N / s, and dynamically adjusts the force distribution ratio on both sides according to the angular velocity during the rotation of the sample tube, effectively solving the problem that traditional constant force grasping is prone to slipping or excessive deformation during rotation operations.

[0050] Please continue to refer to Figure 1 , identify the natural frequency and local deformation characteristics of the object at the grasping force loading position according to the stress wave characteristic mapping data and the grasping strategy data, design an initial force loading sequence, monitor the real-time vibration state and local deformation amount of the object at the grasping force loading position, and dynamically adjust the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data.

[0051] In an embodiment of the present invention, the identifying the natural frequency and local deformation characteristics of the object at the grasping force loading position according to the stress wave characteristic mapping data and the grasping strategy data, designing an initial force loading sequence, monitoring the real-time vibration state and local deformation amount of the object at the grasping force loading position, and dynamically adjusting the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data includes: Analyze the transient response characteristics of the object at the grasping force loading position according to the frequency response characteristics in the stress wave characteristic mapping data and the initial force loading parameters in the grasping strategy data, and calculate the natural frequency distribution and deformation sensitivity coefficient of the grasping force loading position; Design a force loading reference trajectory based on the natural frequency distribution and the deformation sensitivity coefficient, calculate the acceleration compensation coefficient during the force loading process, generate an anti-inertia compensation force sequence according to the acceleration compensation coefficient, and superimpose the anti-inertia compensation force sequence on the initial force loading parameters to generate an initial force loading sequence; Obtain the vibration response data and local deformation data under the action of the force loading according to the initial force loading sequence, calculate the phase offset of the vibration response data and the strain gradient of the deformation data, and generate real-time monitoring data; Calculate the deformation compensation parameters of each grasping force loading position according to the real-time monitoring data and the deformation sensitivity coefficient, adjust the frequency composition of the force loading sequence in intervals, and generate a dynamic force control sequence and a force compensation coefficient; Calculate the force distribution ratio of each grasping force loading position according to the dynamic force control sequence and the force compensation coefficient, and dynamically compensate the force distribution parameters.

[0052] The following is a specific description of the steps involved in the above embodiments: When analyzing the transient response characteristics of an object at the grasping force loading position based on the frequency response characteristics in the stress wave feature mapping data and the initial force loading parameters in the grasping strategy data, a standardized pulse excitation (frequency range 20 - 300 Hz, amplitude 0.2 - 0.5 N) is first applied at the selected grasping force loading position, and a high-speed vibration detection system (including an acceleration sensor array and a high-speed imaging system) is used to record the dynamic response at this position. The transient response characteristics refer to the transient vibration behavior of the object after being subjected to a short-term force excitation, reflecting the inherent dynamic characteristics of the local structure of the object. By applying fast Fourier transform and modal analysis techniques to the measured data, the frequency response function at the grasping force loading position is extracted. According to the amplitude peak position of the frequency response function, the natural frequency distribution at this position is determined, that is, the set of frequencies at which the object is most likely to resonate at this position. At the same time, based on the peak height and full width at half maximum of the frequency response function, the damping ratio and Q factor (quality factor) of the position are calculated, and these parameters reflect the object's ability to absorb and dissipate vibration energy. Combining the frequency response and damping characteristics, the deformation sensitivity coefficient is calculated, which represents the sensitivity of the object to dynamic loads at the force loading position. The higher the deformation sensitivity coefficient, the more likely it is for this position to produce amplified deformation under dynamic loads. Taking food packaging as an example, the natural frequency test of the sealing edge area shows an obvious resonance peak in the range of 65 - 80 Hz, and the deformation sensitivity coefficient is 3 times that of the flat area, which means that this frequency range needs to be particularly avoided during the grasping force loading process to prevent excessive vibration at the packaging edge and cause seal breakage.

[0053] When designing the force loading reference trajectory based on the natural frequency distribution and deformation sensitivity coefficient, the time profile of the force loading is first determined, and the S-shaped loading curve is used instead of the traditional linear loading to avoid the sudden impact at the beginning of the loading stage. The acceleration and deceleration stages of the S-shaped curve account for 25% of the total loading time respectively, and the intermediate uniform speed stage accounts for 50%. This distribution ensures a smooth transition while maintaining reasonable loading efficiency. The parameters of the force loading curve are dynamically adjusted according to the deformation sensitivity coefficient of the object. The higher the sensitivity coefficient, the slower the loading rate, which is usually controlled in the range of 0.5-3N / s. In order to avoid the natural frequency of the object, the loading curve is designed through spectrum optimization to ensure that the main frequency components in the loading process are staggered by at least 10Hz from the natural frequency of the object. The acceleration compensation coefficient during the force loading process is calculated, which indicates the additional compensation force required to offset the inertial effect generated by the object during the accelerated motion. The compensation coefficient calculation takes into account the mass distribution, support state and expected motion trajectory of the object. For medical flexible catheters, when the acceleration of the grasping point reaches 0.5g, the grasping force needs to be increased by about 20% to offset the inertial effect and avoid slippage. The anti-inertia compensation force sequence is generated according to the acceleration compensation coefficient. The sequence is updated in real time with the object's motion state to ensure accurate compensation during the acceleration and deceleration stages. The anti-inertia compensation force sequence is superimposed on the initial force loading parameters to generate a complete initial force loading sequence. This dynamic compensation mechanism significantly improves the grasping stability of deformed objects under non-static conditions, which is particularly important for grasping operations of medical devices that require rapid movement.

[0054] When obtaining the vibration response data and local deformation data under the action of force loading according to the initial force loading sequence, a real-time monitoring system is used to synchronously collect multi-modal feedback information during the grasping process. The vibration response data is collected by a micro-accelerometer (sensitivity 10mg / √Hz, frequency range 5 - 500Hz), and the sampling frequency is set to 1kHz to ensure the capture of high-frequency transient responses. The local deformation data is measured by an optical displacement sensor or a digital image correlation system with an accuracy of 10μm, which can accurately record the micro-deformation process. Signal processing techniques, including band-pass filtering, envelope analysis, and short-time Fourier transform, are applied to the vibration response data to extract the frequency components, amplitude changes, and time evolution characteristics of the vibration. The phase offset of the vibration response data is calculated, and this parameter reflects the time difference between the vibrations of different parts of the object and is a key indicator for identifying structural heterogeneity and potential unstable regions. At the same time, the local deformation data is analyzed, and the strain gradient, that is, the strain change rate per unit distance, is calculated. Regions with too large strain gradients are usually deformation concentration areas and are at risk of structural failure. The real-time monitoring data records the dynamic response of the object during the grasping process with high time resolution, providing timely feedback for subsequent force control. Taking a medical dialysis membrane as an example, real-time monitoring during the force loading process found that the vibration phase in the central region of the membrane always leads the edge region by about 30 degrees, indicating that the central region structure is relatively loose, and the strain gradient reaches a peak near the fixed edge. This information directly guides the adjustment of the spatial distribution of the grasping force to ensure uniform force and avoid the destruction of the microporous structure caused by local overstretching of the dialysis membrane.

[0055] When calculating the deformation compensation parameters for each grasping force loading position based on the real-time monitoring data and the deformation sensitivity coefficient, a closed-loop control algorithm is used to dynamically adjust the grasping strategy according to the actual response of the object monitored. The deformation compensation parameters include the force amplitude compensation coefficient and the force direction compensation vector. The former adjusts the magnitude of the force, and the latter fine-tunes the acting direction of the force. The calculation process is based on the deviation between the measured vibration and deformation data and the expected response, and is weighted and optimized in combination with the deformation sensitivity coefficient of the object. The frequency composition of the force loading sequence is adjusted in intervals. Specifically, through digital filtering technology, the frequency components close to the natural frequency of the object (within the range of ±5Hz) are suppressed, and at the same time, the frequency interval components with good damping characteristics of the object are enhanced to form a dynamic force control sequence with the characteristic of "frequency avoidance". The force compensation coefficient is calculated according to the real-time deformation state of the object. When it is detected that the local deformation exceeds the safety threshold (usually 1.3 times the expected deformation), the force compensation coefficient in the corresponding region is reduced to reduce the local force and prevent excessive deformation. Taking the grasping of a soft robot as an example, real-time monitoring shows that the vibration amplitude at its joints increases non-linearly with the increase of force. When the force reaches 2N, the vibration amplitude suddenly increases by 50%. The system immediately adjusts the force compensation coefficient in the joint area to 0.7 and at the same time increases the suppression component with a frequency of 150Hz, successfully preventing the vibration amplification and ensuring the grasping stability of the flexible joint.

[0056] When calculating the force distribution ratio at each grasping force loading position according to the dynamic force control sequence and the force compensation coefficient, the cooperative control problem of multi-point grasping is considered. For grasping tasks using multiple contact points, the magnitudes and directions of the forces at each contact point need to be coordinated and cooperated to achieve stable grasping together. The calculation of the force distribution ratio is based on the principle of global optimization, and the objective function takes into account three aspects: grasping stability, deformation minimization, and energy efficiency. For typical two-point or three-point grasping, according to the mass distribution and geometric characteristics of the object, an initial force distribution ratio is set (such as a uniform distribution on both sides of the centroid), and then it is dynamically adjusted based on the real-time monitored response of the object. The dynamic compensation of the force distribution parameters adopts real-time feedback control, and the compensation frequency reaches 100 Hz to ensure timely response to changes in the object state. In specific implementation, the control system calculates the deviation between the current force distribution and the target state based on the measured contact force and the object response, and then dynamically adjusts the force control parameters of each contact point according to the magnitude and change trend of the deviation. Taking a medical flexible infusion bag as an example, when grasping the infusion bag for a flipping operation, the system detects an increase in the tangential force caused by the movement of the liquid center of gravity, immediately increases the force of the lower grasping point by 30%, and at the same time adjusts the force direction of the upper grasping point to deflect 15 degrees, effectively offsetting the torque brought by the liquid movement and preventing the infusion bag from slipping during the flipping process. This dynamic adaptability cannot be achieved by traditional fixed-parameter grasping methods.

[0057] In an embodiment of the present invention, designing a force loading reference trajectory based on the natural frequency distribution and the deformation sensitivity coefficient, calculating an acceleration compensation coefficient during the force loading process, generating an anti-inertia compensation force sequence according to the acceleration compensation coefficient, and superimposing the anti-inertia compensation force sequence on the initial force loading parameters to generate an initial force loading sequence includes: Determining the motion constraint space of the robot at each grasping force loading position according to the natural frequency distribution, combining the deformation sensitivity coefficient to calculate the pose change range during the loading process, and generating force loading reference trajectory data; Calculating the joint acceleration distribution of the robot at the key path points according to the force loading reference trajectory data, performing piecewise smoothing processing on the acceleration distribution, and calculating the inertia compensation parameters of each path point based on the smoothed acceleration distribution to generate an acceleration compensation coefficient; Calculating the magnitude and direction of the inertial force at each grasping force loading position according to the acceleration compensation coefficient, generating anti-inertia compensation force data, and performing phase synchronization processing on the anti-inertia compensation force data to generate an anti-inertia compensation force sequence; Superimposing the anti-inertia compensation force sequence on the initial force loading parameters according to the phase superposition rule to generate an initial force loading sequence.

[0058] The following specifically describes the steps involved in the above embodiment: When determining the motion constraint space of the robot at each grasping force loading position according to the natural frequency distribution, first analyze the influence of the natural frequency characteristics of the object on the robot motion. The motion constraint space refers to the speed, acceleration, and frequency limit regions that the robot needs to follow to avoid exciting unstable vibrations of the object when performing the grasping task. In specific implementation, convert the natural frequency distribution data of the object into corresponding time-domain motion constraints, mainly including the maximum allowable acceleration, the maximum motion speed, and the key frequency avoidance interval. For objects with relatively low natural frequencies (such as 10 - 30 Hz), limit the maximum acceleration of the robot's motion, usually set to 0.1 - 0.3 g, to avoid generating excitations close to the natural frequency of the object. For objects with natural frequencies in the intermediate frequency range (30 - 100 Hz), mainly limit the robot's motion rhythm and the contact establishment rate, and control the grasping speed within the range of 2 - 5 mm / s. Combine the deformation sensitivity coefficient of the object to calculate the pose change range during the loading process. The higher the deformation sensitivity coefficient of the region, the smaller the allowable pose change range. For high-sensitivity regions (sensitivity coefficient > 0.8), the pose change is limited within the range of ±1 degree angle and ±0.5 mm. Based on these constraint conditions, use the cubic spline interpolation algorithm to generate a continuous and smooth trajectory from the initial approaching position to the final grasping position to ensure that the entire motion process meets the constraint requirements. Taking the grasping of a medical catheter as an example, there is an obvious resonance in the catheter wall within the range of 45 - 60 Hz. The system automatically limits the frequency components during the robot's acceleration process outside this interval, and strictly limits the pose change near the joint according to the deformation sensitivity coefficients of different sections of the catheter (0.9 at the joint and 0.3 in the homogeneous section) to ensure that the grasping process does not excite unstable vibrations of the catheter.

[0059] When calculating the joint acceleration distribution of the robot at the key path points based on the force loading reference trajectory data, first convert the Cartesian space trajectory of the end effector into a joint space trajectory. Through the inverse kinematics algorithm of the robot, calculate the joint angle configuration corresponding to each key point on the trajectory. Key path points refer to important positions such as direction change points, speed change points, and contact establishment points on the trajectory, and usually about 10 - 20 points are evenly selected throughout the trajectory. For each key path point, calculate the acceleration values of all joints of the robot to form an acceleration distribution matrix. The acceleration is calculated using the central difference method, that is, perform a time difference on the velocities of adjacent path points. Perform piecewise smoothing on the calculated acceleration distribution, and use a low-pass filter algorithm (cutoff frequency set to 10 - 15Hz) to eliminate the high-frequency fluctuation components in the acceleration curve to avoid tremors during robot execution. The smoothing process pays special attention to the acceleration change rate (i.e., jerk), ensuring that it does not exceed the dynamic response ability of the robot execution system, generally controlled within the range of 50 - 100g / s. Based on the smoothed acceleration distribution, calculate the inertial compensation parameters for each path point. The inertial compensation parameters are a set of coefficients used to describe the magnitude of the inertial effect that needs to be compensated when the robot performs acceleration and deceleration motions. The calculation considers factors such as the mass of the object, the contact state between the object and the gripper, and the angle between the acceleration direction and the normal direction of the contact surface. The finally generated acceleration compensation coefficients are stored in the form of a lookup table for easy real-time query and interpolation calculation. For a medical test kit with uneven mass distribution, the acceleration compensation coefficient is about 30% higher on the side of the center of gravity than in the light area. This differential compensation ensures that the test kit does not tilt or slip in the gripper during rapid movement, improving the operation accuracy.

[0060] When calculating the magnitude and direction of the inertial forces at each grasping force loading position according to the acceleration compensation coefficient, Newton's second law is applied, combined with the mass characteristics of the object and the motion state of the robot. Inertial force refers to the inertial effect force generated when an object is accelerating. Its magnitude is proportional to the mass and acceleration of the object, and the direction is opposite to the direction of acceleration. The calculation process first determines the effective mass of the object, considering the mass distribution differences in different directions of the object. Especially for non-uniform objects, an equivalent mass matrix is needed to describe the inertial characteristics in each direction. For each grasping force loading position, a three-dimensional inertial force vector is calculated based on the acceleration vector and equivalent mass at that point. The inertial force calculation formula is F = m·a·k, where m is the effective mass of the object, a is the acceleration vector, and k is the acceleration compensation coefficient. According to the calculation results, complete anti-inertial compensation force data is generated, including the magnitude and direction of the compensation force to be applied at each grasping point at each moment in the time series. Phase synchronization processing is performed on the anti-inertial compensation force data to ensure the coordination of the compensation forces at multiple points during multi-point grasping, and to avoid the cancellation of the compensation forces between different grasping points or the generation of additional torques. Phase synchronization uses time alignment and filtering techniques to make the compensation forces at all grasping points change according to a predetermined phase relationship. For a medical infusion bag containing liquid, the flow of the liquid during rotational motion will cause a lagging inertial effect, and the phase of the compensation force needs to be delayed by about 50 - 100 ms accordingly. This precise phase synchronization processing effectively solves the problem of grasping instability caused by liquid sloshing.

[0061] When superimposing the anti-inertia compensation force sequence and the initial force loading parameters according to the phase superposition rule, it is necessary to consider the force synthesis effect and the mutual influence between different frequency components. The phase superposition rule means maintaining the phase relationship of each component when synthesizing the force signal to ensure that the superposition result meets the expected control effect. In specific implementation, first convert the initial force loading parameters into a time-domain force sequence, and then directly superimpose the anti-inertia compensation force sequence onto the initial force sequence in the time domain. During the superposition process, apply frequency-domain filtering technology to avoid the resonance amplification effect caused by the same frequency components in the two sequences. For the force components near the natural frequency of the object, adopt the phase cancellation technology, that is, ensure that the phase difference between the compensation force and the initial force at these frequencies is 180 degrees to achieve mutual cancellation and reduce the risk of exciting resonance. Normalize the initial force loading sequence generated after the superposition process to ensure that the overall force amplitude meets the grasping requirements while maintaining the original dynamic characteristics. For complex objects with multiple grasping points, such as the actuator of a medical robot, the superposition process will consider the spatial relationship and force transmission path between each point, and use a joint optimization method to calculate the final force loading sequence for the grasping points that are relatively close to avoid local overload. Through this fine force control sequence design, the system can achieve high-precision grasping control of deformable objects, and can maintain stable grasping without causing excessive deformation or unstable vibration of the object even during rapid movement, which is of great significance for the automated operation of deformation-sensitive objects such as medical devices, food packaging, and precision electronic equipment.

[0062] The method for a robot to grasp an object in the embodiments of the present invention has been described above. Next, the device for a robot to grasp an object in the embodiments of the present invention will be described. Please refer to Figure 2 , an embodiment of the device for a robot to grasp an object in the embodiments of the present invention includes: A feature mapping module 101, configured to obtain the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object by performing multi-point pulse excitation on the object, perform time-frequency analysis on the obtained data, extract the parameters of the stress wave propagating inside the object, and generate stress wave feature mapping data including the stress propagation speed distribution, vibration response amplitude, and frequency response characteristics; A structure analysis module 102, configured to analyze the propagation characteristics of the stress wave inside the object according to the stress wave feature mapping data, identify the stress transmission channels, determine the position of the material boundary and the deformation-sensitive area inside the object, and generate internal structure characteristic data including the stiffness distribution, stress transmission channels, and deformation-sensitive area distribution; A grasping planning module 103, configured to calculate the deformation risk coefficient and stability contribution degree of each area on the surface of the object according to the internal structure characteristic data, select the grasping positions that are closely connected to the stress transmission channels and avoid the deformation-sensitive areas, and generate grasping strategy data including the grasping force loading positions and the initial force loading parameters; The force control execution module 104 is configured to identify the natural frequency and local deformation characteristics of an object at the grasping force loading position according to the stress wave feature mapping data and the grasping strategy data, design an initial force loading sequence, monitor the real-time vibration state and local deformation amount of the object at the grasping force loading position, and dynamically adjust the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data.

[0063] above Figure 2 The device for a robot to grasp an object in the embodiments of the present invention is described in detail from the perspective of modular functional entities. Next, the device for a robot to grasp an object in the embodiments of the present invention is described in detail from the perspective of hardware processing.

[0064] Figure 3 FIG. is a schematic structural diagram of a device for a robot to grasp an object provided by an embodiment of the present invention. The device 200 for a robot to grasp an object may vary greatly due to different configurations or performances, and may include one or more processors 210 (for example, one or more processors) and a memory 220, and one or more storage media 230 (for example, one or more mass storage device ends) storing application programs 233 or data 232. Among them, the memory 220 and the storage media 230 may be transient storage or persistent storage. The program stored in the storage media 230 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the device 200 for a robot to grasp an object. Further, the processor 210 may be configured to communicate with the storage media 230 and execute a series of instruction operations in the storage media 230 on the device 200 for a robot to grasp an object to implement the steps of the method for a robot to grasp an object as described above.

[0065] The device 200 for a robot to grasp an object may further include one or more power supplies 240, one or more wired or wireless network interfaces 250, one or more input / output interfaces 260, and / or one or more operating systems 231, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, and so on. Those skilled in the art can understand that Figure 3 The shown structural diagram of the device for a robot to grasp an object does not limit the device for a robot to grasp an object provided by the present invention, and may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0066] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are run on a computer, the computer is caused to execute the steps of the method for the robot to grasp an object.

[0067] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, device, or unit can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0068] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc that can store program codes.

[0069] The foregoing are only the preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural transformation made by using the description and drawings of the present invention under the inventive concept of the present invention, or direct / indirect application in other related technical fields, is included in the patent protection scope of the present invention.

Claims

1. A method for a robot to grasp an object, characterized in that, Including: By performing multi-point pulse excitation on an object, acquiring the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object, conducting time-frequency analysis on the acquired data, extracting the parameters of stress wave propagation inside the object, and generating stress wave characteristic mapping data including stress propagation velocity distribution, vibration response amplitude, and frequency response characteristics; According to the stress wave characteristic mapping data, analyzing the propagation characteristics of stress waves inside the object, identifying stress transmission channels, determining the material boundary positions and deformation-sensitive regions inside the object, and generating internal structure characteristic data including stiffness distribution, stress transmission channels, and deformation-sensitive region distribution; According to the internal structure characteristic data, calculating the deformation risk coefficients and stability contribution degrees of each region on the object surface, selecting the grasping positions that are closely connected to the stress transmission channels and avoiding the deformation-sensitive regions, and generating grasping strategy data including the grasping force loading positions and initial force loading parameters; According to the stress wave characteristic mapping data and the grasping strategy data, identifying the natural frequencies and local deformation characteristics of the object at the grasping force loading positions, designing the initial force loading sequence, monitoring the real-time vibration state and local deformation amount of the object at the grasping force loading positions, and dynamically adjusting the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data.

2. The method for a robot to grasp an object according to claim 1, wherein, The step of "By performing multi-point pulse excitation on an object, acquiring the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object, conducting time-frequency analysis on the acquired data, extracting the parameters of stress wave propagation inside the object, and generating stress wave characteristic mapping data including stress propagation velocity distribution, vibration response amplitude, and frequency response characteristics" includes: Selecting multiple pulse excitation points on the object surface according to the spatial coverage principle, and applying a pulse sequence with a frequency range of 5 - 500 Hz and an amplitude of 0.1 - 0.5 N to the pulse excitation points; Acquiring the vibration acceleration data of the object surface according to the pulse sequence, performing Fourier transform on the vibration acceleration data to obtain the spectrum data of each excitation point, and calculating the vibration response amplitude according to the spectrum data; Acquiring the displacement change data of the object surface according to the pulse sequence, performing wavelet decomposition on the displacement change data to obtain the surface transient deformation data, and analyzing the deformation response of the object surface according to the surface transient deformation data; Acquiring the three-dimensional point cloud data of the object according to the pulse sequence, calculating the deformation amount of the object surface to obtain the three-dimensional deformation data; Performing time-frequency joint analysis on the vibration response amplitude, surface transient deformation data, and three-dimensional deformation data, and extracting the velocity parameters, energy attenuation parameters, and phase parameters of stress wave propagation inside the object; Generating the stress propagation velocity distribution according to the velocity parameters, calculating the vibration response amplitude according to the energy attenuation parameters and phase parameters, analyzing the frequency response characteristics according to the spectrum data of each excitation point, and generating the stress wave characteristic mapping data.

3. The method for a robot to grasp an object according to claim 1, wherein, Analyze the propagation characteristics of stress waves inside the object based on the stress wave feature mapping data, identify the stress transmission channels, determine the positions of material boundaries and deformation-sensitive regions inside the object, and generate internal structure characteristic data including stiffness distribution, stress transmission channels, and deformation-sensitive region distribution, including: Calculate the mechanical impedance coefficients of each region of the object based on the stress propagation velocity distribution, analyze the material boundary surfaces inside the object based on the mechanical impedance coefficients, calculate the deformation response parameters of the material boundary surfaces in combination with the vibration response amplitude and frequency response characteristics, and determine the material boundary position data; Apply a preloading force for robot grasping to the object surface according to the stress propagation velocity distribution and the vibration response amplitude, analyze the propagation path of stress waves under the preloading force, identify the bifurcation and convergence nodes and the main transmission channels of stress waves inside the object, and generate stress transmission channel data; Calculate the dynamic stiffness coefficients of each region inside the object based on the stress transmission channel data and the frequency response characteristics, analyze the distribution law of the dynamic stiffness coefficients along the stress transmission channels, and generate stiffness distribution data; Conduct a correlation analysis on the stiffness distribution data, stress transmission channel data, and material boundary position data, calculate the deformation response distribution of the object in the robot grasping direction, divide the deformation-sensitive regions based on the deformation response distribution, and generate deformation-sensitive region distribution data; Calculate the deformation transfer coefficient of the object under the action of robot grasping based on the stiffness distribution data, stress transmission channel data, and deformation-sensitive region distribution data, and generate internal structure characteristic data.

4. The method for a robot to grasp an object according to claim 3, wherein The step of applying a preloading force for robot grasping to the object surface according to the stress propagation velocity distribution and the vibration response amplitude, analyzing the propagation path of stress waves under the preloading force, identifying the bifurcation and convergence nodes and the main transmission channels of stress waves inside the object, and generating stress transmission channel data includes: Calculate the preloading contact area on the object surface according to the stress propagation velocity distribution, generate a preloading depth gradient map for the preloading contact area, and plan the robot preloading motion trajectory in combination with the vibration response amplitude; Apply a preloading force to the object surface along the preloading motion trajectory, obtain the contact force feedback data during the preloading process, adjust the action direction and depth of the preloading force according to the contact force feedback data, and generate preloading force distribution data; Analyze the stress propagation wavefront inside the object based on the preloading force distribution data, identify the wavefront bifurcation points and wavefront convergence points, calculate the attenuation rate of wavefront propagation in combination with the vibration response amplitude, and generate stress wave propagation path data; Conduct a connectivity analysis on the stress wave propagation path data, select the propagation paths with an attenuation rate lower than a preset threshold as the main transmission channels, and generate stress transmission channel data.

5. The method for a robot to grasp an object according to claim 1, characterized in that, Calculate the deformation risk coefficients and stability contribution degrees of each region on the object surface based on the internal structure characteristic data, select the grasping positions that are closely connected to the stress transmission channels and avoid the deformation-sensitive regions, and generate grasping strategy data including the grasping force loading positions and initial force loading parameters, including: Calculate the deformation response range of each area on the object surface in the grasping direction according to the stiffness distribution data and the deformation-sensitive area distribution data of the internal structure characteristic data, calculate the deformation transfer coefficient of each area in combination with the stress transfer channel, and generate a deformation risk coefficient based on the deformation response range and the deformation transfer coefficient; Calculate the control ability of each area on the object surface for stress transfer according to the stress transfer channel data and the stiffness distribution data of the internal structure characteristic data, and generate a stability contribution degree based on the control ability; According to the deformation risk coefficient and the stability contribution degree, screen candidate grasping areas outside the distribution of the deformation-sensitive areas, calculate the connection node density between the candidate grasping areas and the stress transfer channel, and select the area where the connection node density meets the grasping stability requirement as the grasping force loading position; Calculate the acting direction and force loading amplitude of the grasping force according to the stiffness distribution of the grasping force loading position, and generate initial force loading parameters; Generate grasping strategy data according to the grasping force loading position and the initial force loading parameters.

6. The method for a robot to grasp an object according to claim 1, wherein According to the stress wave characteristic mapping data and the grasping strategy data, identify the natural frequency and local deformation characteristics of the object at the grasping force loading position, design an initial force loading sequence, monitor the real-time vibration state and local deformation amount at the grasping force loading position of the object, and dynamically adjust the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data, including: Analyze the transient response characteristics of the object at the grasping force loading position according to the frequency response characteristics in the stress wave characteristic mapping data and the initial force loading parameters in the grasping strategy data, and calculate the natural frequency distribution and deformation sensitivity coefficient of the grasping force loading position; Design a force loading reference trajectory based on the natural frequency distribution and the deformation sensitivity coefficient, calculate the acceleration compensation coefficient during the force loading process, generate an anti-inertia compensation force sequence according to the acceleration compensation coefficient, and superimpose the anti-inertia compensation force sequence on the initial force loading parameters to generate an initial force loading sequence; According to the initial force loading sequence, obtain the vibration response data and local deformation data under the action of the force loading, calculate the phase offset of the vibration response data and the strain gradient of the deformation data, and generate real-time monitoring data; Calculate the deformation compensation parameters of each grasping force loading position according to the real-time monitoring data and the deformation sensitivity coefficient, perform interval adjustment on the frequency composition of the force loading sequence, and generate a dynamic force control sequence and a force compensation coefficient; Calculate the force distribution ratio of each grasping force loading position according to the dynamic force control sequence and the force compensation coefficient, and perform dynamic compensation on the force distribution parameters.

7. The method for a robot to grasp an object according to claim 6, wherein, The design of the force loading reference trajectory based on the natural frequency distribution and the deformation sensitivity coefficient, the calculation of the acceleration compensation coefficient during the force loading process, the generation of the anti-inertia compensation force sequence according to the acceleration compensation coefficient, and the superimposition of the anti-inertia compensation force sequence on the initial force loading parameters to generate an initial force loading sequence include: Determine the motion constraint space of the robot at each grasping force loading position according to the natural frequency distribution, calculate the pose change range during the loading process in combination with the deformation sensitivity coefficient, and generate force loading reference trajectory data; Calculate the joint acceleration distribution of the robot at the key path points according to the force loading reference trajectory data, perform piecewise smoothing processing on the acceleration distribution, calculate the inertial compensation parameters of each path point based on the smoothed acceleration distribution, and generate acceleration compensation coefficients; Calculate the magnitude and direction of the inertial force at each grasping force loading position according to the acceleration compensation coefficients, generate anti-inertial compensation force data, and perform phase synchronization processing on the anti-inertial compensation force data to generate an anti-inertial compensation force sequence; Superimpose the anti-inertial compensation force sequence and the initial force loading parameters according to the phase superposition rule to generate an initial force loading sequence.

8. A device for a robot to grasp an object, characterized in that, The device for the robot to grasp an object adopts the method for the robot to grasp an object according to any one of claims 1 to 7. The device for the robot to grasp an object includes: A feature mapping module, configured to obtain the surface vibration data, surface transient deformation data, and three-dimensional deformation data of the object by performing multi-point pulse excitation on the object, perform time-frequency analysis on the obtained data, extract the parameters of the stress wave propagating inside the object, and generate stress wave feature mapping data including the stress propagation speed distribution, vibration response amplitude, and frequency response characteristics; A structure analysis module, configured to analyze the propagation characteristics of the stress wave inside the object according to the stress wave feature mapping data, identify the stress transmission channels, determine the material boundary positions and deformation sensitive regions inside the object, and generate internal structure characteristic data including the stiffness distribution, stress transmission channels, and deformation sensitive region distribution; A grasping planning module, configured to calculate the deformation risk coefficient and stability contribution degree of each region on the object surface according to the internal structure characteristic data, select the grasping positions that are tightly connected to the stress transmission channels and avoid the deformation sensitive regions, and generate grasping strategy data including the grasping force loading positions and initial force loading parameters; A force control execution module, configured to identify the natural frequency and local deformation characteristics of the object at the grasping force loading position according to the stress wave feature mapping data and the grasping strategy data, design an initial force loading sequence, monitor the real-time vibration state and local deformation amount of the object at the grasping force loading position, and dynamically adjust the frequency composition and force distribution parameters of the force loading sequence according to the monitoring data.

9. A device for a robot to grasp an object, characterized in that, The device for the robot to grasp an object includes: a memory and at least one processor, and instructions are stored in the memory; The at least one processor calls the instructions in the memory so that the device for the robot to grasp an object executes the steps of the method for the robot to grasp an object according to any one of claims 1-7.

10. A computer-readable storage medium, on which instructions are stored, characterized in that, When the instructions are executed by the processor, the steps of the method for the robot to grasp an object according to any one of claims 1-7 are implemented.

Images