Fluid-driven micro-manipulator gripping device

Abstract

The application discloses a fluid driving-based micro-manipulator clamping device and belongs to the technical field of mechanical arms, comprising a large arm and a small arm; the end of the small arm is provided with a clamping part and a control unit, the clamping part is composed of two clamping jaws, the clamping jaws are flexible, and the control unit performs the following steps to realize the grasping control of a target object: acquiring fluid instantaneous temperature, instantaneous flow and instantaneous ion conductivity, constructing a fluid medium characteristic model to output fluid medium characteristic coefficients; acquiring driving pressure high-frequency pulsation root mean square, base vibration amplitude and actively applied high-frequency vibration acceleration, constructing a dynamic interference model to output dynamic interference coefficients; the fluid cavitation or clamping jaw paralysis failure risk caused by outputting illegal negative absolute pressure instructions is completely eliminated, and the absolute safety and stability of bottom-layer operation under complex working conditions are greatly ensured.

Description

Technical Field

[0001] This invention relates to the field of robotic arm technology, and in particular to a fluid-driven micro-manipulator gripping device. Background Technology

[0002] Existing fluid-driven robotic arms typically employ traditional linear closed-loop feedback (such as PID) control strategies, relying primarily on the single absolute pressure feedback of the fluid loop, combined with basic vision or macroscopic tactile sensors to roughly determine the grasping state. In terms of the actuation and perception models, existing solutions mostly assume that the actuating fluid medium is always in an ideal steady state and only focus on the overall macroscopic deformation of the robotic arm, usually idealizing the contact interface between the gripper and the target object as a static, absolutely rigid coupling.

[0003] However, the aforementioned existing technologies have significant drawbacks in real-world complex operating conditions: First, they do not consider the nonlinear attenuation effect of dynamic deviations in the fluid medium itself (such as temperature, flow rate, and conductivity) on the underlying drive efficiency; second, when faced with vibrations of the robotic arm base or high-frequency fluid pulsation, they lack a dynamic disturbance quantification model, which can easily lead to severe system oscillations; third, existing sensing systems suffer from severe misalignment between microscopic transient quantities and macroscopic state quantities, making it impossible to accurately extract the transient micro-slip rate of the interface and the maximum principal strain in the characteristic region, thus making it difficult to truly characterize the fluid-structure interaction degree; finally, under extreme nonlinear disturbances, traditional pressure control algorithms are prone to crossing physical boundaries and falling into mathematical singularities, outputting illegal negative absolute pressure commands, leading to fluid cavitation or gripper paralysis failure.

[0004] Therefore, the present invention provides a fluid-driven micro manipulator gripping device. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by the present invention to solve its technical problem is: a fluid-driven micro manipulator gripping device, comprising an upper arm and a forearm; The forearm end is equipped with a gripping part and a control unit. The gripping part consists of two flexible grippers. The control unit performs the following steps to achieve gripping control of the target object: Obtain the instantaneous temperature, instantaneous flow rate, and instantaneous ionic conductivity of the fluid, construct a fluid medium characteristic model, and output the fluid medium characteristic coefficients; The root mean square of the high-frequency pulsation of the driving pressure, the vibration amplitude of the base, and the actively applied high-frequency vibration acceleration are obtained to construct a dynamic disturbance model and output the dynamic disturbance coefficient. Obtain the spatial distribution variance of clamping contact force, the transient micro-slip rate of the interface, and the actual surface contact angle; construct an interface contact quality model and output the interface contact quality coefficient. By combining the fluid medium characteristic coefficient and dynamic disturbance coefficient, and obtaining the maximum principal strain and absolute displacement of the claw in the characteristic region, a fluid-structure interaction synergy model is constructed to output the synergy degree. Based on the degree of cooperation, the interface contact quality coefficient, and the obtained current absolute pressure of the driving fluid, a pressure optimization model is constructed to output the target absolute pressure of the driving fluid to drive the control gripper.

[0007] The beneficial effects of this invention are as follows: 1. The fluid-driven micro-manipulator gripping device of this invention effectively overcomes the inherent defects of traditional linear closed-loop feedback control strategies by setting an outermost physical extreme value cutoff protection mechanism for the fluid medium characteristic model and pressure optimization model. It not only acquires the instantaneous temperature, flow rate and ionic conductivity of the fluid in real time, quantifies and compensates for the nonlinear attenuation effect of the dynamic deviation of the medium itself on the underlying driving efficiency; but also compares the corrected theoretical target driving pressure with the preset safe pressure holding baseline pressure through maximum value function logic, and selects the larger one for output, completely eliminating the risk of fluid cavitation or gripper paralysis failure caused by outputting illegal negative absolute pressure commands, and greatly ensuring the absolute safety and stability of the underlying operation under complex working conditions.

[0008] 2. The fluid-driven micro-manipulator gripping device of the present invention, by setting up a multi-source dynamic interference monitoring architecture including a high-frequency dynamic piezoresistive pressure sensor, a triaxial accelerometer, and a micro piezoelectric accelerometer, constructs a dynamic interference model that can output smooth dimensionless coefficients. It overcomes the limitation of existing technologies in lacking quantitative analysis of environmental disturbances, accurately extracts the root mean square of high-frequency pulsation of driving pressure, the amplitude of base vibration, and the actively applied high-frequency vibration acceleration, endowing the control system with excellent anti-interference ability, effectively eliminating the system oscillation that is easily caused by high-frequency fluid pulsation or severe base oscillation, and significantly improving the adaptive performance of the equipment in dynamic and harsh environments.

[0009] 3. The fluid-driven micro-manipulator gripping device described in this invention achieves fine control of the gripping state by setting up a multi-dimensional microscopic sensing system composed of a tactile sensing array and a strain sensor network, as well as a corresponding interface contact quality model and fluid-structure interaction synergy model. It successfully solves the industry problem of severe misalignment between microscopic transient quantities and macroscopic state quantities, and greatly enhances the fit and reliability of the gripper when gripping small and fragile objects. Attached Figure Description

[0010] Figure 1 This is a perspective view of the present invention; Figure 2 This is a flowchart of the overall grasping control process in this invention; Figure 3This is a flowchart of the fluid medium characteristic model in this invention; Figure 4 This is a flowchart of the dynamic interference model in this invention; Figure 5 This is a flowchart of the interface contact quality model in this invention; Figure 6 This is a flowchart of the fluid-structure interaction synergy model in this invention; Figure 7 This is a flowchart of the pressure optimization model in this invention.

[0011] In the diagram: 1. Upper arm; 2. Forearm; 3. Clamping part; 4. Gripper. Detailed Implementation

[0012] The present invention provides a fluid-driven micro-manipulator gripping device, comprising an upper arm 1 and a forearm 2. The forearm 2 is equipped with a gripping part 3 and a control unit at its end. The gripping part 3 consists of two grippers 4, which are flexible. The control unit performs the following steps to achieve gripping control of the target object: Obtain the instantaneous temperature, instantaneous flow rate, and instantaneous ionic conductivity of the fluid, construct a fluid medium characteristic model, and output the fluid medium characteristic coefficients; The root mean square of the high-frequency pulsation of the driving pressure, the vibration amplitude of the base, and the actively applied high-frequency vibration acceleration are obtained to construct a dynamic disturbance model and output the dynamic disturbance coefficient. Obtain the spatial distribution variance of clamping contact force, the transient micro-slip rate of the interface, and the actual surface contact angle; construct an interface contact quality model and output the interface contact quality coefficient. By combining the fluid medium characteristic coefficient and dynamic disturbance coefficient, and obtaining the maximum principal strain and absolute displacement of the tip of the gripper 4 in the characteristic region, a fluid-structure interaction synergy model is constructed to output the synergy degree. Based on the degree of cooperation, the interface contact quality coefficient, and the obtained current absolute pressure of the driving fluid, a pressure optimization model is constructed to output the target absolute pressure of the driving fluid to drive the control gripper 4.

[0013] The gripper 4 is flexible, and the gripping part 3, as the actuator, directly contacts the target object. The flexibility of the gripper 4 allows it to adapt to objects of different shapes and sizes, achieving compliant gripping. The control unit, as the core processing unit, is responsible for coordinating and managing the entire gripping process. For example, the robotic arm body can adopt a multi-joint structure to provide a flexible range of motion; the gripping part 3 can be quickly replaced through modular design; the gripper 4 can be made of elastic materials such as silicone rubber, and its deformation is achieved by filling and releasing fluid in the internal fluid chamber; the control unit can be an embedded system that runs a real-time operating system and control algorithms.

[0014] To assess the state of the driving fluid in real time, the control unit acquires the fluid's instantaneous temperature, instantaneous flow rate, and instantaneous ionic conductivity. Based on these parameters, it constructs a fluid medium characteristic model to output a fluid medium characteristic coefficient. The instantaneous fluid temperature can be measured using a thermistor or platinum resistance temperature sensor. The instantaneous flow rate can be detected using a turbine flow meter or ultrasonic flow meter, and the instantaneous ionic conductivity can be acquired using a conductivity probe. This real-time data reflects the physicochemical properties of the fluid medium. For example, an increase in fluid temperature may lead to a decrease in viscosity, flow rate fluctuations may affect the drive response speed, and changes in ionic conductivity may indicate fluid contamination or aging. The fluid medium characteristic model integrates these raw measurements to calculate a dimensionless fluid medium characteristic coefficient, which quantifies the impact of the current state of the fluid medium on drive performance.

[0015] To quantify the dynamic disturbances experienced by the system, the control unit acquires the root mean square of the high-frequency pulsation of the driving pressure, the amplitude of the base vibration, and the actively applied high-frequency vibration acceleration. Based on these parameters, a dynamic disturbance model is constructed to output a dynamic disturbance coefficient. The root mean square of the high-frequency pulsation of the driving pressure can be obtained by detecting the driving pressure signal through a high-frequency response piezoelectric sensor or piezoresistive sensor and performing time-domain analysis. The amplitude of the base vibration can be measured by a triaxial accelerometer installed on the base of the robotic arm. The actively applied high-frequency vibration acceleration can be detected by a miniature accelerometer integrated on the gripper 4. These parameters can capture dynamic disturbances from the pump system, the external environment, or internal vibration sources. The dynamic disturbance model integrates and processes this multi-source disturbance information to output a dimensionless dynamic disturbance coefficient, which reflects the degree of dynamic instability currently faced by the system.

[0016] To accurately characterize the contact state between the gripper 4 and the target object, the control unit acquires the spatial distribution variance of the gripping contact force, the transient micro-slip rate of the interface, and the actual surface contact angle. Based on these parameters, an interface contact quality model is constructed to output the interface contact quality coefficient. The spatial distribution variance of the gripping contact force can be obtained by integrating a tactile sensor array on the surface of the gripper 4 and statistically analyzing the force values ​​at each sensing point in the array. The transient micro-slip rate of the interface... This can be obtained by analyzing the high-frequency fluctuation characteristics of the tangential force detected by the tactile sensing array. The specific calculation logic is as follows: The control unit first extracts the small high-frequency oscillation amplitude of the tangential force through the tactile sensing array. Combined with the pre-calibrated local Coulomb friction coefficient between the surface of the gripper 4 and the target object and the system's equivalent mass / impedance model, the dynamic change of the tangential force is converted into the local transient relative acceleration of the contact interface. Subsequently, the relative acceleration signal is integrated in the time domain within a short high-frequency passband to obtain the accurate transient micro-slip rate of the interface. .

[0017] The actual contact angle of the surface can be obtained by capturing images of the contact area between the gripper 4 and the object using a miniature vision acquisition device, and then performing image processing and geometric calculations. These parameters provide detailed information about the microscopic behavior of the gripping interface. For example, an excessively large variance in the contact force distribution may indicate uneven gripping, a non-zero micro-slip rate indicates the occurrence of slippage, and a deviation of the actual contact angle from the ideal value may indicate insufficient coverage. The interface contact quality model integrates these microscopic contact parameters and outputs a dimensionless interface contact quality coefficient, which quantifies the effectiveness and stability of the gripping contact.

[0018] To comprehensively evaluate the coordination between fluid drive and mechanical structure response, the control unit combines the aforementioned fluid medium characteristic coefficient and dynamic disturbance coefficient, and obtains the maximum principal strain and tip absolute displacement of the characteristic region of gripper 4. It then constructs a fluid-structure interaction coordination degree model to output the coordination degree. The maximum principal strain of the characteristic region of gripper 4 can be monitored in real time by arranging a strain sensor network in the key deformation area of ​​gripper 4. The tip absolute displacement can be obtained through an optical tracking system or by solving based on the kinematic model of the robotic arm. The fluid medium characteristic coefficient and dynamic disturbance coefficient provide information on fluid drive and environmental disturbance, while the maximum principal strain and tip absolute displacement reflect the actual deformation response of gripper 4. The fluid-structure interaction coordination degree model integrates these multi-dimensional information to output a dimensionless coordination degree. This coefficient comprehensively evaluates the matching degree between fluid drive, environmental disturbance, and the deformation of gripper 4 itself, thus reflecting the overall coordination of the system operation.

[0019] To achieve precise drive control of gripper 4, the control unit constructs a pressure optimization model based on the aforementioned coordination degree, interface contact quality coefficient, and the acquired current absolute pressure of the driving fluid to output a target absolute pressure of the driving fluid. The current absolute pressure of the driving fluid can be detected in real time by an absolute pressure sensor installed on the main fluid loop. The coordination degree and interface contact quality coefficient provide comprehensive feedback on the overall system performance and gripping contact quality. The pressure optimization model uses this real-time feedback information as input to dynamically calculate and output an optimized target absolute pressure of the driving fluid. For example, if the coordination degree or interface contact quality coefficient is low, the model may adjust the target pressure to enhance the gripping force or improve the contact state. This model aims to ensure stable and reliable gripping under various working conditions by adaptively adjusting the driving pressure, while avoiding physical singularities or illegal negative pressure commands that may occur in traditional control methods. The output target absolute pressure of the driving fluid is then sent to the fluid drive actuator to precisely control the movement of gripper 4.

[0020] Traditional fluid-driven robotic arms, when grasping small and fragile components, typically ignore the nonlinear attenuation of underlying actuation performance caused by dynamic deviations in fluid medium properties such as temperature, flow rate, and conductivity. For example, in the example above, when fluid temperature increases leading to a decrease in viscosity, traditional solutions cannot detect and compensate for this change, potentially causing a deviation between the actual grasping force and the expected value. This application, by acquiring instantaneous fluid temperature, instantaneous flow rate, and instantaneous ionic conductivity, and constructing a fluid medium characteristic model to output fluid medium characteristic coefficients, can quantify and compensate for the impact of fluctuations in fluid physical properties on actuation performance in real time, ensuring the accuracy of the actuation force output and thus avoiding grasping instability caused by changes in fluid medium properties.

[0021] By constructing a multi-dimensional perception and control model, a comprehensive quantification of fluid medium characteristics, external environmental interference, and interface contact quality was achieved. Based on this, fluid-structure interaction synergy evaluation and pressure optimization were carried out, thereby accurately controlling the driving pressure. This effectively improved the grasping stability and reliability of the micro-manipulator in complex dynamic environments, solved many problems existing in the prior art, and demonstrated significant technological progress.

[0022] The fluid medium characteristic model calculates the dimensionless deviation between the instantaneous temperature, instantaneous flow rate, and instantaneous ionic conductivity of the fluid and their corresponding preset reference values. Combined with their respective preset sensitivity weights, it adopts a negative correlation nonlinear decay mapping relationship to output fluid medium characteristic coefficients with values ​​greater than zero and less than or equal to one. The more the instantaneous parameters deviate from the reference values, the closer the fluid medium characteristic coefficients are to zero.

[0023] The fluid medium property model is as follows:

[0024] in, The fluid medium property coefficient is dimensionless and has a range of . ; and These are the instantaneous fluid temperature obtained through a temperature sensor and the preset reference temperature, respectively, with dimensions of... ; and These are the instantaneous flow rate obtained through the flow meter and the preset rated reference flow rate, respectively, with dimensions of... ; and These are the instantaneous ionic conductivity obtained through a conductivity probe and the preset reference conductivity, respectively, with dimensions of... ; All are preset dimensionless sensitivity weighting coefficients.

[0025] Among them, fluid medium characteristic coefficient It is a dimensionless quantity with a range of . The coefficient is used to quantify the degree of deviation of the actual performance or health state of a fluid medium from its ideal state. The closer the coefficient is to 1, the closer the fluid medium is to its ideal state; conversely, the smaller the coefficient is, the greater the deviation and the lower its driving efficiency may be.

[0026] Fluid instantaneous temperature This refers to the actual temperature of a fluid measured in real time by a temperature sensor at any given moment. This temperature sensor can be of various types, such as a thermistor, thermocouple, or platinum resistance thermometer, to ensure measurement accuracy and response speed, with a preset reference temperature. It is the ideal or rated operating temperature determined by the system design or fluid material specifications. It is usually stored and retrieved as a fixed parameter in the control unit, and its value can be determined by experimental calibration or according to the fluid property manual.

[0027] Instantaneous flow This refers to the volumetric flow rate of fluid per unit time, which is obtained in real time by a flow meter. Flow meters can be selected from miniature turbine flow meters, ultrasonic flow meters, or Coriolis flow meters, etc., to meet the special requirements of miniature robotic arms for flow measurement accuracy and volume. The preset rated reference flow rate... This is the flow rate value that the system should maintain under ideal operating conditions, also stored in the control unit, serving as a benchmark for measuring the degree of deviation from instantaneous flow rate. Instantaneous ionic conductivity. Conductivity refers to the fluid conductivity measured in real time by a conductivity probe, which reflects the ion concentration or impurity content in the fluid. Conductivity probes can be two-electrode or four-electrode sensors to adapt to different fluid-mediated environments.

[0028] Preset reference conductivity It is the conductivity value of a fluid in a pure or ideal state, serving as a reference benchmark for judging the purity or stability of a fluid's composition. (Sensitivity weighting coefficient) These are dimensionless preset parameters used to adjust the model's sensitivity to deviations of the three physical quantities—temperature, flow rate, and ionic conductivity—from the baseline values. These weighting coefficients can be set through experimental optimization or expert experience based on the fluid medium's response characteristics to changes in different parameters, ensuring that the model accurately reflects the actual impact of each parameter on fluid drive performance.

[0029] By constructing a fluid medium property model based on Gaussian exponential decay, a quantitative characterization of the deviation of the fluid's physical state was achieved. The core logic of this model lies in utilizing instantaneous temperature. Instantaneous flow and instantaneous ionic conductivity With their respective preset reference temperatures Rated reference flow rate and reference conductivity The sum of squared deviations was used to construct a comprehensive index reflecting the degree to which the fluid medium deviates from its ideal state. This was achieved by introducing a temperature sensor, flow meter, and conductivity probe to monitor the instantaneous temperature of the fluid. Instantaneous flow and instantaneous ionic conductivity This scheme enables real-time monitoring and can dynamically capture the nonlinear changes of the fluid medium during operation. The exponential decay function used in the model ensures that the fluid medium characteristic coefficients decrease as the fluid parameters deviate from the preset benchmark. It will exhibit a non-linear decrease, thus being able to sensitively reflect the negative impact of fluid medium property fluctuations on drive performance.

[0030] By setting sensitivity weighting coefficients This scheme endows the system with the ability to adjust the sensitivity to fluctuations in different physical parameters, enabling the model to be adapted differently according to the characteristics of specific fluid media. This method of mapping multidimensional physical parameters into a single dimensionless coefficient provides accurate medium state input for subsequent fluid-structure interaction synergy models and pressure optimization models, effectively solving the problem of inaccurate drive control caused by ignoring the dynamic characteristics of the fluid medium.

[0031] By using the fluid medium characteristic coefficient As a key input parameter of the fluid-structure interaction synergy model, this scheme enables the control unit of the micro-manipulator gripper to more comprehensively evaluate the actual performance of the current fluid drive. Thus, when constructing the pressure optimization model, it can calculate the absolute pressure of the target driving fluid based on a more realistic fluid state, thereby achieving precise drive control of the gripper 4.

[0032] The dynamic disturbance model divides the root mean square of the high-frequency pulsation of the driving pressure, the amplitude of the base vibration, and the actively applied high-frequency vibration acceleration by their respective preset safety thresholds for dimensionless processing. It then performs a weighted summation by combining the preset influence weights of different disturbance sources. After deducting the preset system steady-state bias constant, it outputs the dynamic disturbance coefficient in the open interval from zero to one through an S-shaped nonlinear activation function with smooth saturation characteristics.

[0033] The dynamic interference model is as follows:

[0034] in, The dynamic interference coefficient is dimensionless and has a range of [value missing]. ; and These are the root mean square of the high-frequency pulsation of the driving pressure extracted by the dynamic pressure sensor and the preset nominal safety threshold pressure, respectively, with dimensions of... ; and These are the base vibration amplitude obtained through the first accelerometer and the preset maximum allowable vibration amplitude, respectively, with dimensions of... ; and These are the actively applied high-frequency vibration acceleration and the maximum rated acceleration, respectively, obtained through the second accelerometer, with dimensions of... ; The dimensionless influence weights for different interference sources; and These are the dimensionless decay slope and steady-state bias constant, respectively.

[0035] Among them, the dynamic disturbance model aims to quantify and characterize the impact of various disturbance factors on the system performance of fluid-driven micro-manipulators in complex dynamic environments. Its core function is to integrate multiple physical disturbance quantities (such as pressure pulsation, base vibration, and active vibration) into a unified, dimensionless dynamic disturbance coefficient so that the control system can comprehensively evaluate and compensate for them. This model can be constructed by means of empirical data fitting, physical mechanism modeling, or hybrid modeling. Its output dynamic disturbance coefficient reflects the degree of adverse effect of the current environment on clamping stability.

[0036] Dynamic interference coefficient This is the output of the dynamic disturbance model, used to characterize the overall degree of dynamic disturbance experienced by the system under the current environment, and its value range is... Generally, the closer the value is to 1, the greater the interference and the more significant the impact on system stability; the closer the value is to 0, the smaller the interference and the more stable the system operation. This coefficient is dimensionless, which makes it easy to process in a unified manner with the coefficients output by other models.

[0037] High-frequency pulsation root mean square of driving pressure This refers to the root mean square (RMS) value of the high-frequency pressure fluctuation signal caused by factors such as pump source, valve switching, and pipeline resonance in the driving fluid loop. It quantifies the dynamic instability inside the fluid drive system. This value can be obtained by installing a high-frequency dynamic pressure sensor in the fluid loop, performing bandpass filtering on the collected pressure time-domain signal, and then calculating its RMS value. Another method is to use fast Fourier transform to analyze the spectrum of the pressure signal, extract the energy of the high-frequency components, and calculate its equivalent RMS value.

[0038] Preset nominal safety threshold pressure This is a reference pressure value set during system design based on the structural strength of the robotic arm, the pressure resistance of the fluid circuit, and the desired stable operating conditions. It represents the upper limit that the root mean square of the high-frequency pulsation of the driving pressure should not exceed within the normal or safe operating range. This value is usually determined through experimental testing, simulation analysis, or according to industry standards, and is used to measure the actual measured root mean square of the high-frequency pulsation of the driving pressure. Normalization is performed.

[0039] Base vibration amplitude This refers to the maximum displacement of the vibration occurring at the base where the robotic arm body connects to the external environment. This vibration may originate from the operation of a workbench, production line, or other external equipment. This value can be obtained by installing an accelerometer at the base of the robotic arm, performing a second integration on the collected acceleration signal to obtain the displacement signal, and extracting its peak value or root mean square value as the vibration amplitude. Alternatively, it can be obtained by directly measuring the displacement change of the base using a laser displacement sensor.

[0040] Preset maximum permissible vibration amplitude This value, set during system design, represents the maximum vibration displacement that the robotic arm base can withstand without affecting gripping accuracy and stability. Exceeding this threshold will significantly degrade system performance or even cause failure. This value is typically determined based on the robotic arm's structural characteristics, the robustness of the control algorithm, and the accuracy requirements of the application scenario. It is used to measure the actual vibration amplitude of the base. Normalize.

[0041] Actively applied high-frequency vibration acceleration This refers to the acceleration generated by high-frequency vibrations actively applied to the gripper 4 by the system to achieve certain specific functions (such as assisted gripping, desorption, cleaning, etc.). This vibration is typically generated by a piezoelectric actuator, ultrasonic transducer, or miniature vibration motor. This value can be obtained by integrating a miniature accelerometer onto the gripper 4 to directly measure its acceleration response under active vibration excitation. Alternatively, if the characteristics of the active vibration source are known and stable, the generated acceleration can be estimated using control commands or a preset model. Maximum rated acceleration. This value is set during system design and represents the maximum design acceleration that gripper 4 can achieve when actively applying high-frequency vibration. It represents the upper limit of the performance of the active vibration source and also serves as a reference for the actual measured acceleration of actively applied high-frequency vibration. The benchmark for normalization. This value is usually determined by the performance specifications of the active vibration actuator.

[0042] Dimensionless influence weights of different interference sources This is used to adjust the relative importance of different disturbance sources (high-frequency pulsation of driving pressure, base vibration, and actively applied high-frequency vibration) in the dynamic disturbance model. For example, if the driving pressure pulsation has the greatest impact on the system, then... The values ​​of these weights will be relatively large. These weights are usually determined through experimental calibration, expert experience, or optimization algorithms so that the dynamic disturbance coefficients output by the model can most accurately reflect the actual disturbance situation.

[0043] attenuation slope Used to adjust the steepness of the Sigmoid function, thereby controlling the dynamic disturbance coefficient. Sensitivity to changes in input variables is relatively high. A higher value makes the function curve steeper, meaning the system is more sensitive to disturbances. When the disturbance reaches a certain level, the dynamic disturbance coefficient will increase rapidly. The value makes the curve flatter and the system response relatively sluggish. This value is usually optimized through system identification or empirical adjustment.

[0044] steady-state bias constant Used to adjust the horizontal offset of the Sigmoid function, thereby controlling the dynamic disturbance coefficient. The initial output level when the input variable is zero or a small value allows the model to output a baseline dynamic disturbance coefficient even without significant disturbances, or to adjust the output to the desired range at a specific disturbance level. This value is typically set experimentally or empirically.

[0045] Dynamic pressure sensors are used to monitor pressure changes in fluid loops in real time, especially high-frequency fluctuations. They can be piezoresistive, piezoelectric, or capacitive types, featuring high response frequencies and high sensitivity, capable of capturing microsecond-level pressure transients. Their installation position is typically near the drive fluid inlet of the gripper 4 to directly reflect the fluid pressure state acting on the gripper 4. The first accelerometer measures the vibration of the robotic arm base, usually a triaxial accelerometer capable of simultaneously measuring acceleration in the X, Y, and Z directions, thus comprehensively reflecting the base's vibration state. It can be a MEMS accelerometer or a piezoelectric accelerometer, installed at the connection between the robotic arm body and the fixed base to accurately capture vibrations transmitted from the external environment. The second accelerometer measures the active vibration acceleration of the gripper 4 itself. Since the gripper 4 is typically small, a miniature, lightweight accelerometer, such as a MEMS accelerometer or a miniature piezoelectric accelerometer, is required. It is usually integrated into the base of the gripper 4 or a key stress area to accurately monitor the dynamic response of the gripper 4 under active excitation.

[0046] Dynamic pressure sensors extract the root mean square of high-frequency driving pressure pulsations in real time. and the preset nominal safety threshold pressure Ratio normalization is performed to quantify the impact of high-frequency pressure fluctuations within the fluid circuit caused by pump source or pipeline effects on clamping stability. Simultaneously, the first accelerometer acquires the vibration amplitude of the base. and the preset maximum allowable vibration amplitude Ratio normalization is performed to monitor the degree of disturbance of the robotic arm body by external environmental vibration in real time. In addition, a second accelerometer acquires the actively applied high-frequency vibration acceleration. and with maximum rated acceleration Ratio normalization is performed to incorporate vibration disturbances introduced by active control into the model. The three normalized disturbance indices are then weighted according to their respective dimensionless influence weights. Perform a weighted summation and combine it with the decay slope. With steady-state bias constant The input is fed into the Sigmoid function, and the final output is the dynamic interference coefficient. This approach transforms complex physical disturbances into smooth, dimensionless coefficients, enabling the control system to handle disturbances from different sources using a unified mathematical language. This dynamic disturbance coefficient... The fluid medium characteristic coefficients output by the above fluid medium characteristic model can be compared with those of the fluid medium characteristic model. The synergistic effect provides a more comprehensive and accurate system state assessment for the subsequent fluid-structure interaction synergy model, so that when driving and controlling the gripper 4, not only the characteristics of the fluid itself are considered, but also the dynamic disturbances of the external environment and internal drive are fully considered.

[0047] The interface contact quality model extracts the spatial distribution variance of clamping contact force, the transient micro-slip rate of the interface, and the actual contact angle of the surface. It calculates the dimensionless relative deviations between these deviations and the preset non-uniformity tolerance variance threshold, the maximum allowable slip rate, and the optimal theoretical covering contact angle. It assigns a corresponding scaling factor to each deviation and performs a fusion mapping through a centrally symmetric bounded nonlinear translation function to output the interface contact quality coefficient with a value range in the open interval from zero to one.

[0048] The interface contact quality model is as follows:

[0049] in, The interface contact quality coefficient is dimensionless and has a range of [value missing]. ; and These are the spatial distribution variance of the clamping contact force obtained through the tactile sensor array and the preset non-uniformity tolerance variance threshold, respectively, with dimensions of... ; and These are the acquired transient micro-slip rate of the interface and the preset maximum allowable slip rate, respectively, with dimensions of... ; and These are the actual surface contact angle obtained through a visual acquisition device and the preset optimal theoretical coverage contact angle, respectively, with dimensions of... ; All are dimensionless scaling factors.

[0050] An interface contact quality model is a mathematical model used to quantify the quality or stability of the contact interface between a micro-manipulator gripper 4 and a target object. It integrates multiple physical quantities related to the contact state and outputs a single, dimensionless coefficient to characterize the quality of the contact. This model can be constructed based on empirical data fitting, physical principle derivation, or machine learning methods. For example, a nonlinear function (such as the hyperbolic tangent function) can be used to map multiple input parameters to an output value between 0 and 1, where 1 represents ideal contact and 0 represents complete contact failure. Another implementation approach is to use a fuzzy logic system or neural network to weightedly fuse inputs from different sensors, outputting a comprehensive contact quality assessment.

[0051] Spatial distribution variance of clamping contact force Variance is a statistical measure of the uniformity of force distribution on the contact surface between the gripper 4 and the target object. The larger the variance, the more uneven the contact force distribution, and there may be local overpressure or underpressure areas. The smaller the variance, the more uniform the force distribution and the more stable the grip. This variance can be obtained by acquiring discrete force values ​​on the contact surface through the tactile sensor array, and then calculating the variance of these force values. For example, the output force value of each sensing unit of the tactile sensor array can be used as a sample to calculate its statistical variance. Another way to achieve this is to perform a two-dimensional Gaussian fitting on the force distribution of the tactile sensor array and extract the variance of the fitting residual as an indicator of non-uniformity.

[0052] Non-uniformity tolerance variance threshold This is a preset reference value used to determine whether the gripping contact force distribution is within an acceptable uniform range. This threshold can be obtained through experimental calibration. For example, under different gripping conditions, a maximum tolerable variance value can be determined by statistically analyzing the force distribution variance of a large number of successful gripping cases. Another approach is to set an upper limit for the force distribution variance within a safe range based on the mechanical properties of the gripper 4 material, the fragility of the target object, and the requirements of the gripping task, through simulation or theoretical calculation. A tactile sensor array is a sensor system composed of multiple independent tactile sensor units that can sense the pressure distribution, shear force, or vibration information at the contact interface between the gripper 4 and the target object in real time. Tactile sensor arrays can be implemented using technologies such as piezoresistive sensors, capacitive sensors, or piezoelectric sensors. For example, multiple miniature piezoresistive sensors can be integrated on a flexible substrate to form a two-dimensional array, with each sensor unit outputting a resistance change signal proportional to the local pressure. Another approach is to use a distributed tactile sensor based on a fiber Bragg grating (FBG) to indirectly obtain the contact force distribution by measuring the fiber strain.

[0053] Interface transient micro-slip rate This refers to the relative displacement velocity that occurs at the contact interface between the gripper 4 and the target object within a very short time. Micro-slippage is a precursor to macroscopic slippage of the object during the gripping process, and its real-time monitoring is crucial for maintaining gripping stability. This rate can be obtained by time-domain integration of the high-frequency fluctuation characteristics of the tangential force detected by the tactile sensing array. For example, when the object undergoes micro-slippage, the tangential force at the contact interface will generate high-frequency oscillations. By demodulating and integrating these oscillation signals, the transient slippage rate can be estimated. Another approach is to integrate a miniature inertial measurement unit (IMU) or optical flow sensor on the surface of the gripper 4 to directly measure the local relative motion velocity of the contact area.

[0054] Maximum permissible slip ratio This is a preset threshold value, representing the maximum transient micro-slip rate that the interface between the gripper 4 and the target object can tolerate during the grasping process. This threshold value can be set based on the surface friction characteristics of the target object, the material of the gripper 4, the accuracy requirements of the grasping task, and experimental data. For example, by conducting a slip experiment under controlled conditions, the maximum micro-slip rate before the object begins macroscopic slippage is recorded, and a safety margin is set based on this.

[0055] Actual contact angle of surface This refers to the actual enveloping angle formed between the flexible surface of the gripper 4 and the surface of the target object when the gripper 4 grasps the target object. This angle reflects the degree of contact and integrity of the gripper 4 with the object. This angle can be obtained by acquiring images of the gripper 4 and the target object through a visual acquisition device (such as a camera), and then using image processing algorithms (such as edge detection and contour fitting) to calculate the actual contact angle between the surface of the gripper 4 and the surface of the object. For example, by performing three-dimensional reconstruction on the grasped image, and then measuring the angle between the normal vectors of the contact area between the gripper 4 and the object. Another way to achieve this is to embed a flexible position sensor or strain sensor inside the gripper 4, and indirectly infer its enveloping angle with the object by the deformation of the gripper 4.

[0056] Optimal theoretical coverage contact angle This is a preset ideal angle, representing the theoretical contact angle at which the gripper 4 achieves the most stable and effective grasping of the target object. This angle can be determined through theoretical calculations or simulations based on the structural design of the gripper 4, the geometric model of the target object, and the optimization objectives of the grasping task. When facing a target object of unknown shape or variable size, the control unit acquires the 3D point cloud contour of the target object in real time through a vision acquisition device. Combined with the flexible kinematic model of the gripper 4, it iteratively calculates online the dynamic characteristic angle that maximizes the interface coverage contact area, and uses this as the optimal theoretical coverage contact angle under the current working conditions. In the input interface contact quality model.

[0057] For example, for objects of a specific shape, the covering angle at which the gripper 4 achieves the maximum contact area or the most uniform stress distribution can be determined through finite element analysis or geometric modeling. A vision acquisition device is a device that can acquire image or video data to monitor the macroscopic contact state and geometric relationship between the gripper 4 and the target object. The vision acquisition device may include, but is not limited to, industrial cameras, miniature cameras, depth cameras or stereo vision systems. For example, a high-resolution miniature camera can be installed at the end of a robotic arm or near the gripper 4 to capture contact images between the gripper 4 and the target object in real time.

[0058] Dimensionless scaling factor These are weighting parameters used to adjust the contribution of various input terms (such as force distribution uniformity, slip ratio, and contact angle) in the interface contact quality model to the final contact quality coefficient. They ensure that different physical quantities have appropriate influence in the model. These factors can be determined through experimental calibration, optimization algorithms, or expert experience. For example, by conducting a series of grasping experiments, recording the impact of different parameter changes on the grasping success rate, and then using regression analysis or genetic algorithms to optimize these scaling factors so that the model output best matches the actual grasping performance.

[0059] hyperbolic tangent function It is an S-shaped function with a range of -1 to 1. It is often used to map any real number input to a finite output range and has good smoothness and nonlinearity. In the interface contact quality model, the hyperbolic tangent function is used to map the combined result of multiple weighted input terms to a contact quality coefficient of 0 to 1.

[0060] The above solution achieves a refined quantitative assessment of the contact state between the micro-manipulator gripper 4 and the target object by constructing an interface contact quality model. This model comprehensively considers the variance of the spatial distribution of the gripping contact force. , Interface transient micro-slip rate and the actual contact angle of the surface These three key metrics, and utilize a dimensionless scaling factor. These factors are weighted, and the final result is mapped to an interface contact quality coefficient between 0 and 1 using a hyperbolic tangent function. .

[0061] Specifically, the variance of the spatial distribution of clamping contact force Data is acquired in real time through a tactile sensing array and compared with a preset non-uniformity tolerance variance threshold. A comparison is made to assess the uniformity of the contact force. When the actual variance is much smaller than the threshold, it indicates that the force distribution is highly uniform and contributes positively to the contact quality; conversely, if the variance is too large, it indicates that the force distribution is uneven and may lead to unstable gripping.

[0062] Interface transient micro-slip rate The high-frequency fluctuation characteristics of the tangential force detected by the tactile sensing array are obtained by time-domain integration and compared with the preset maximum allowable slip ratio. By comparing the relative motion trends of the contact interface in real time, a low micro-slip rate indicates stable contact and a positive contribution to contact quality; if the slip rate is close to or exceeds the maximum allowable value, it indicates that the gripping is about to fail.

[0063] Actual contact angle of surface Acquired through a visual acquisition device and compared with the preset optimal theoretical coverage contact angle. The comparison quantifies the degree of gripper 4's coverage of the target object. When the actual contact angle is close to the optimal theoretical contact angle, it indicates good coverage and a positive contribution to contact quality. If the deviation is large, it indicates that the coverage is not ideal. These three indicators, after their respective normalization and weighting, are input into the hyperbolic tangent function to generate the final interface contact quality coefficient. This coefficient can intuitively reflect the stability, reliability and fit of the current gripping interface. By transforming the complex physical contact state into a single quantitative indicator, it provides an accurate decision basis for subsequent driving pressure optimization. This interface contact quality model, combined with the above-mentioned fluid medium characteristic model and dynamic disturbance model, forms a more comprehensive gripping state evaluation system.

[0064] The fluid medium property model provides information on the health status of the driving fluid itself, the dynamic disturbance model quantifies the disturbances of the external environment and the system itself, and the interface contact quality model directly reflects the quality of the microscopic interaction between the gripper 4 and the target object. These three elements work together to enable the system not only to sense changes in the driving source and the external environment but also to deeply understand the actual condition of the gripping interface. This allows for more precise adjustment of the driving strategy, avoiding gripping failures due to insufficient information in a single dimension. For example, even if the fluid medium properties are good and external disturbances are minimal, if the interface contact quality coefficient is low (e.g., due to micro-slippage or poor coating), the system can promptly identify and adjust the driving pressure to improve gripping robustness. This multi-dimensional, collaborative sensing strategy significantly enhances the intelligence and reliability of the micro-manipulator's gripping control under complex working conditions.

[0065] The fluid-structure interaction synergy model takes the ratio of the maximum principal strain in the characteristic region to the preset ultimate tensile strain rate, the fluid medium characteristic coefficient, the anti-interference margin after deducting the dynamic interference coefficient, and the penalty term for the absolute displacement of the tip deviating from the target optimal coverage expected displacement as independent dimensions. It performs global situation fusion through a multivariate nonlinear product structure based on the preset elastic exponents of each dimension, and obtains the synergy in the range of zero to one, which is left-closed and right-open, by limiting the output boundary through an exponential saturation function.

[0066] The fluid-structure interaction synergy model is as follows:

[0067] in, For the degree of synergy, dimensionless, with a range of . ; and These are the maximum principal strain and the preset ultimate tensile strain rate of the characteristic region obtained through the strain sensor network, respectively, and are dimensionless. and These are the absolute tip displacement obtained through kinematic calculations and the expected optimal coverage displacement of the current target, respectively, with dimensions of... ; It is a dimensionless elasticity index; is a dimensionless saturation scaling constant.

[0068] Coordination It is an indicator that quantifies the comprehensive coordination of fluid actuation, structural response, external disturbances, and interaction with the target object during the grasping process of a fluid-driven micromanipulator. Its purpose is to provide a unified and comprehensive indicator to evaluate the quality of the grasping state, including the maximum principal strain in the characteristic region. The ultimate tensile strain rate refers to the maximum tensile deformation that the material in a specific region of the gripper 4 can withstand when it is subjected to force and deformation. It reflects the local stress condition and deformation limit of the gripper 4 and is a key parameter for evaluating the structural safety and gripping force of the gripper 4. This parameter can be monitored and extracted in real time through a strain sensor network integrated on the surface of the gripper 4, such as a resistance strain gauge array or a fiber Bragg grating sensor array. This refers to the maximum tensile strain that the material of gripper 4 can withstand without permanent damage or failure. It serves as a safety threshold to regulate the deformation range of gripper 4 and prevent overload damage. Its value can be pre-calibrated through material mechanics experiments or determined according to the mechanical property manual of the gripper 4 material.

[0069] absolute displacement of the tip This refers to the change in spatial position of the tip of gripper 4 relative to the robotic arm body or a fixed reference coordinate system. It reflects the macroscopic motion state of gripper 4 and the degree to which it covers the target object. This displacement can be calculated by combining the encoder of the robotic arm body with the kinematic model of gripper 4, or by real-time measurement through an external visual tracking system, such as a high-speed camera combined with image processing algorithms.

[0070] The current optimal expected displacement for target coverage is the ideal spatial position that the tip of the gripper 4 should reach, which is preset based on the geometry, size, and desired gripping posture of the target object. This position serves as the target for gripping control, guiding the gripper 4 to achieve the best coverage effect. This expected displacement can be calculated based on the target object's CAD model using an offline path planning algorithm, or dynamically adjusted through online visual recognition and target posture estimation.

[0071] Elasticity Index These are dimensionless weighting factors used to adjust the degree of influence of various input parameters in the model on the degree of synergy. By adjusting these indices, the importance of different physical quantities in the synergy assessment can be nonlinearly weighted to adapt to different grasping tasks and environmental conditions. Their values ​​can be set through experimental data fitting, machine learning algorithm training, or expert experience.

[0072] Saturation scaling constant It is a dimensionless constant used to scale the composite terms within the exponential function as a whole, thereby adjusting the response sensitivity and saturation characteristics of the synergy model and ensuring synergy. The range of values ​​is Between these parameters, the rate of change of the model's output under different input conditions is controlled to prevent the model from being too sensitive or insensitive to extreme inputs. The value of this parameter can be determined through system identification, optimization algorithms, or empirical settings.

[0073] By constructing a fluid-structure interaction synergy model, multi-dimensional integration of fluid medium properties, dynamic disturbances, structural strain response, and kinematic displacement deviation is achieved, enabling a comprehensive quantitative evaluation of the robotic arm's grasping state. This model first introduces fluid medium property coefficients. The power function term is used to weight the nonlinear effect on driving performance, ensuring that the model can adjust the synergy evaluation benchmark according to real-time changes in the fluid state. For example, when the fluid medium characteristic coefficient... A lower value indicates poor fluid medium conditions; even with ideal conditions, the degree of synergy will decrease accordingly, reflecting a reduction in actual driving efficiency. Secondly, by introducing a dynamic disturbance coefficient... complement power function term This transforms the negative impact of external vibrations or pressure pulsations on the system into a suppression factor in the synergy assessment. This means that when the dynamic disturbance coefficient... At higher levels, the degree of synergy decreases significantly, allowing the model to effectively eliminate environmental noise interference and reflect the true level of driving synergy. Furthermore, by introducing the maximum principal strain in the feature region... With ultimate tensile strain rate Residual amount penalty item The micro-deformation state of gripper 4 is incorporated into the evaluation system. This penalty term reflects the residual stress tolerance of the gripper 4 material, optimizing the gripping force while ensuring structural safety. For example, when... Increase and approach the ultimate tensile strain rate When the residual penalty term approaches 0, the result of the external negative exponential function approaches 0, and the final synergy will be significantly reduced, accurately indicating that the system gripper 4 has approached its deformation limit.

[0074] Furthermore, by introducing tip absolute displacement The optimal coverage displacement of the current target The deviation exponent term uses a Gaussian function to penalize displacement deviation, ensuring that the robotic arm can accurately conform to the surface of the target object during grasping, achieving optimal coverage. and When the deviation is large, this term will significantly reduce the degree of coordination, prompting the system to adjust to reduce the displacement error. Ultimately, this is achieved through the saturation scaling constant. By performing a nonlinear mapping on the aforementioned multiple coupling factors, complex physical quantities are transformed into a unified synergy index. This provides precise feedback signals for subsequent pressure optimization control. This multi-dimensional, nonlinear fusion evaluation mechanism overcomes the limitations of traditional single pressure feedback, enabling the system to more accurately understand and respond to complex fluid-structure interaction behaviors, thereby achieving smarter and more robust grasping control.

[0075] The pressure optimization model is based on the preset ratio of the desired synergy to the desired synergy and the preset ratio of the desired contact quality to the interface contact quality coefficient. It calculates the corresponding feedback compensation terms through a logarithmic proportional function and combines them with a preset gain coefficient to dynamically and proportionally adjust the current driving fluid absolute pressure. The outermost layer of the pressure optimization model includes a cutoff protection mechanism based on physical extreme values, which ensures that the output target driving fluid absolute pressure is always not lower than the preset safe pressure baseline pressure.

[0076] To avoid pressure output falling into physical singularities and illegal negative pressure ranges, the pressure optimization model is as follows:

[0077] in, The output is the absolute pressure of the target driving fluid, with dimensions of ; The system's preset safety pressure baseline pressure, with dimensions of ; To obtain the current absolute pressure of the driving fluid, the dimension is... ; and These are the preset target expected synergy and expected contact quality, respectively, and are dimensionless. and These are the preset logarithmic feedback gain coefficients for synergy and contact quality, respectively, and are dimensionless. This is a dimensionless minimum regularization constant used to prevent the denominator from being zero.

[0078] The target driving fluid absolute pressure in this model It is the fluid pressure command that is finally output by the control unit to drive the micro-manipulator gripper 4. Its function is to precisely adjust the gripping force or deformation of the gripper 4 to achieve stable and reliable clamping of the target object. The pressure can be positive or negative (vacuum), depending on the design of the gripper 4 and the gripping requirements. For example, the fluid pressure can be controlled by adjusting the opening of the proportional valve.

[0079] Safety holding pressure baseline pressure It is a minimum safe absolute pressure value preset by the system to ensure that the output drive pressure will not be lower than a physically acceptable lower limit under any circumstances. It is usually set to atmospheric pressure or slightly higher than atmospheric pressure to prevent fluid cavitation and thus protect the normal operation of the fluid circuit and gripper 4.

[0080] Current driving fluid absolute pressure It is the actual absolute pressure in the fluid circuit acting on the gripper 4, which is acquired in real time by a sensor. It is the reference point for the system to perform feedback control and reflects the current driving state of the gripper 4. For example, it can be measured in real time by an absolute pressure sensor installed on the main channel of the fluid circuit.

[0081] Preset target expected synergy It is the target value of fluid-structure interaction synergy that the system is expected to achieve under ideal grasping conditions. This value is usually determined during the system design or experimental calibration stage and represents the best grasping performance. For example, an optimal synergy threshold can be determined by simulation analysis or by the stability of gripper 4 and the grasping success rate in a standard grasping experiment.

[0082] Preset expected contact quality This is the target value for interface contact quality that the system expects to achieve under ideal grasping conditions. This value is also determined during the system design or experimental calibration phase and represents the optimal contact state. For example, a contact quality target value that can guarantee stable grasping can be determined by testing different grasping tasks and combining information such as tactile feedback and visual feedback. The preset cooperative degree logarithmic feedback gain coefficient... It is a dimensionless coefficient used to adjust the degree of influence of the coordination deviation on the target driving pressure correction. The selection of this coefficient affects the system's response speed and stability. For example, it can be empirically adjusted through PID controller parameter tuning methods or optimized online through adaptive control algorithms.

[0083] Preset contact quality logarithmic feedback gain coefficient It is a dimensionless coefficient used to adjust the degree of influence of contact quality deviation on the target driving pressure correction. The selection of this coefficient also affects the system's response speed and stability. For example, its optimal value can be determined by system identification methods, or it can be manually adjusted according to the sensitivity of the actual grasping task.

[0084] A dimensionless minimum regularization constant used to prevent the denominator from being zero. It is a very small positive number added to the denominator of the logarithmic function to prevent a mathematical division-by-zero error when the real-time value of synergy or contact quality approaches zero. For example, it can be set to an extremely small positive floating-point number, such as... or To ensure the continuity of the calculation, the maximum value function logic used in the model is used to compare the calculated theoretical target driving pressure with the preset safe holding pressure baseline pressure. The larger of the two values ​​is selected as the final output pressure. Its core function is to ensure that the output driving pressure is always not lower than the safety threshold, thereby avoiding physical singularity problems such as negative pressure command and fluid cavitation. For example, in software implementation, the maximum value function provided by the programming language can be called directly.

[0085] In pressure optimization models, logarithmic functions are used to convert the deviation between coordination degree and contact quality into a correction amount for driving pressure. The nonlinear characteristics of logarithmic functions enable the system to have high sensitivity when the error is large, allowing it to respond and correct quickly, while exhibiting stable adjustment capability when the error is small, which helps to avoid overshoot and oscillation. For example, logarithmic functions can be calculated using mathematical libraries.

[0086] By integrating multi-dimensional information and employing a nonlinear feedback mechanism, precise and robust pressure control of a fluid-driven micro-manipulator gripper was achieved. The model first obtains the current absolute pressure of the driving fluid. It serves as a control benchmark. Based on this, it utilizes a pre-set target expectation degree of synergy. With desired contact quality The degree of synergy is calculated in real time using the fluid-structure interaction synergy model and the interface contact quality model. and interface contact quality coefficient Compare them.

[0087] When real-time coordination or interface contact quality coefficient When the value deviates from its expected value, the model introduces a logarithmic feedback gain coefficient of synergy. and contact quality logarithmic feedback gain coefficient Using a logarithmic function to measure the absolute pressure of the current driving fluid Nonlinear corrections are performed, leveraging the properties of the logarithmic function to provide significant correction when the system deviates greatly from the target, while fine-tuning occurs when approaching the target. This avoids the oscillations or sluggish response that can occur with traditional linear feedback. To ensure numerical stability, a minimal regularization constant is introduced into the denominator of the logarithmic function. This effectively prevents division-by-zero errors that may occur when the degree of coordination or contact quality approaches zero. Finally, to avoid the pressure output falling into physical singularities and illegal negative pressure ranges, the model adopts maximum value function logic, which compares the modified theoretical target driving pressure with the system's preset safe pressure holding baseline pressure. The values ​​are compared, and the larger one is selected as the target absolute pressure of the driving fluid for the final output. To prevent the system from falling into an open-loop runaway state due to forced cutoff when the theoretical target driving pressure remains consistently low, the control unit is equipped with a follow-up compensation strategy: when the safe pressure baseline is output for multiple consecutive control cycles... When the system is in operation, the control unit will trigger system-level collaborative compensation, actively outputting commands to make the robotic arm base fine-tune the grasping posture to improve the contact quality of the physical interface, or directly trigger an audible and visual alarm to terminate the current grasping cycle, ensuring that the system maintains the underlying logical closed loop while avoiding negative pressure singularities.

[0088] This overall mechanism ensures that even in complex dynamic environments, such as changes in fluid medium properties, external dynamic disturbances, and uncertainties in the microscopic state of the clamping interface, the system can output physically feasible and effective pressure commands to drive the gripper 4, thereby maintaining a stable gripping state and avoiding the risk of fluid cavitation or gripper 4 failure.

[0089] The fluid circuit of the driving gripper 4 integrates a temperature sensor for acquiring instantaneous temperature, a flow meter for acquiring instantaneous flow rate, and a conductivity probe for acquiring instantaneous ionic conductivity. The main channel of the fluid circuit is equipped with an absolute pressure sensor to detect the current absolute pressure of the driving fluid, and a high-frequency dynamic piezoresistive pressure sensor to extract the fluid time-domain signal and calculate the root mean square of the high-frequency pulsation of the driving pressure.

[0090] Temperature sensors are used to monitor the instantaneous temperature of fluid media in real time. These sensors can be implemented using various technologies. For example, a thermistor can be used, whose resistance changes significantly with temperature, and the temperature can be calculated by measuring the resistance. Alternatively, a thermocouple can be used, which uses the Seebeck effect to convert the temperature difference into a potential difference for measurement.

[0091] Flow meters are used to measure the instantaneous flow rate of fluid media in real time. They can be implemented in ways including but not limited to turbine flow meters, which calculate the flow rate by measuring the rotational speed of the turbine driven by the fluid; or, ultrasonic flow meters, which determine the flow velocity and then calculate the flow rate by measuring the time difference of ultrasonic waves propagating in the fluid.

[0092] Conductivity probes are used to acquire the instantaneous ionic conductivity of fluid media in real time. These probes are typically composed of electrodes and reflect the ion concentration of the fluid by measuring the conductivity between the electrodes. For example, two-electrode or four-electrode conductivity sensors can be used.

[0093] An absolute pressure sensor is used to detect the current absolute pressure of the driving fluid in a fluid circuit. This sensor can provide static pressure information for the fluid circuit and can be implemented as a piezoresistive sensor, which uses the piezoresistive effect of a semiconductor material to convert pressure into an electrical signal; or it can use a capacitive sensor, which determines the pressure by measuring the change in capacitance caused by pressure.

[0094] High-frequency dynamic piezoresistive pressure sensors are specifically designed to capture high-frequency pressure fluctuation signals in fluid circuits and calculate the root mean square of high-frequency driving pressure pulsations based on these signals. Unlike conventional pressure sensors, these sensors have higher response frequencies and sensitivity, enabling them to accurately reflect fluid pulsation characteristics. For example, MEMS (Micro-Electro-Mechanical Systems) piezoresistive pressure sensors can be used, whose miniature structure and high bandwidth characteristics enable them to effectively capture rapidly changing pressure signals.

[0095] By integrating temperature sensors, flow meters, and conductivity probes into the fluid loop, the system can monitor the temperature, flow rate, and ionic conductivity of the fluid medium in real time. These parameters are the core inputs for constructing the fluid medium characteristic model, enabling the control unit to dynamically adjust the drive strategy according to the real-time characteristics of the fluid medium, thereby compensating for the impact of changes in fluid medium characteristics on drive performance. An absolute pressure sensor is installed in the main channel of the fluid loop to directly obtain the current absolute pressure of the drive fluid. This parameter serves as the benchmark input for the pressure optimization model, ensuring the accuracy of the target drive pressure calculation and avoiding control failures caused by pressure sensing deviations. By setting up a high-frequency dynamic piezoresistive pressure sensor, the system can capture high-frequency time-domain signals in the fluid circuit and calculate the root mean square of the high-frequency pulsation of the driving pressure. This characteristic parameter is an important component of the dynamic disturbance model, enabling the system to quantify the disturbance caused by fluid pulsation. This allows the introduction of a corresponding compensation mechanism during pressure optimization, effectively improving the gripping stability of the manipulator in complex dynamic environments. This comprehensive sensor integration solution enables the control unit to acquire rich and accurate real-time data that is difficult to achieve with traditional solutions. This allows for a more accurate assessment of fluid medium characteristics, dynamic disturbances, and the current driving pressure, providing solid data support for subsequent fluid-structure interaction synergy models and pressure optimization models. This significantly improves the gripping accuracy and stability of the micro manipulator under complex working conditions.

[0096] A triaxial accelerometer is installed at the base of the robotic arm to obtain the vibration amplitude of the base. A piezoelectric accelerometer is integrated at the bottom of the gripper 4 to detect the actively applied high-frequency vibration acceleration. A tactile sensor array and a strain sensor network are integrated on the surface of the gripper 4. The control unit performs time-domain integration based on the high-frequency fluctuation characteristics of the tangential force detected by the tactile sensor array to obtain the transient micro-slip rate of the interface, and extracts the maximum tensile principal strain value in the area covered by the strain sensor network as the maximum principal strain of the feature region.

[0097] A triaxial accelerometer is a sensor that can simultaneously measure the acceleration of an object in three orthogonal directions, inferring the object's motion state by detecting inertial forces. In a miniature robotic gripper, this accelerometer is mounted on the base of the robotic arm to monitor the base's vibration in real time. For example, an accelerometer based on microelectromechanical systems (MEMS) technology can be used, which is small in size, low in power consumption, and suitable for integration; or a piezoelectric accelerometer can be used, which measures acceleration by utilizing the principle that piezoelectric materials generate charges when deformed under force. By processing the triaxial acceleration signals, the vibration amplitude of the base can be calculated. This provides key inputs for subsequent dynamic disturbance models.

[0098] A piezoelectric accelerometer is a sensor that converts mechanical vibration into an electrical signal using the piezoelectric effect. It is particularly suitable for detecting high-frequency vibrations. Its principle is that when a piezoelectric material is subjected to acceleration, it generates an electric charge proportional to the acceleration. In the clamping device, this piezoelectric accelerometer is integrated into the base of the gripper 4, specifically for detecting actively applied high-frequency vibration acceleration. For example, a shear-type piezoelectric accelerometer can be used, which is insensitive to lateral vibration and can more accurately measure acceleration in a specific direction. This sensor can precisely capture the dynamic response of gripper 4 under an active vibration strategy, providing real-time feedback to the control system.

[0099] A tactile sensor array is a planar array composed of multiple independent tactile sensor units. It can sense information such as pressure distribution, shear force, and texture on the surface of an object. Its concept is to simulate the tactile function of biological skin and provide precise contact perception. Integrating the tactile sensor array on the surface of the gripper 4 can acquire the variance of the spatial distribution of the clamping contact force in real time. The system detects high-frequency fluctuations in tangential force. For example, a piezoresistive sensor array can be used, sensing force through changes in resistance of conductive rubber or thin films under pressure; or a capacitance-based sensor array can be used, sensing force through changes in capacitance caused by deformation of an elastic medium. These arrays provide high spatial resolution mechanical information, which is key to understanding the interface interaction between the gripper 4 and the target object. A strain sensor network is a system composed of multiple strain sensors (such as strain gauges) connected in a specific layout to measure local deformation or strain on the object's surface. The concept is to infer the internal stress state of a material by measuring changes in its geometric dimensions under force. Integrating a strain sensor network on the surface of the gripper 4 allows monitoring of deformation in different areas of the gripper 4 during the gripping process. For example, a traditional metal foil strain gauge can be used to measure strain through changes in resistance; or a fiber Bragg grating (FBG) sensor can be used to measure strain through spectral drift caused by periodic changes in the grating in the fiber, offering advantages such as resistance to electromagnetic interference. This network provides detailed data on the deformation of the gripper 4, which is an important basis for evaluating the fluid-structure interaction synergy.

[0100] The control unit performs time-domain integration based on the high-frequency fluctuation characteristics of tangential force detected by the tactile sensor array to obtain the transient micro-slip rate of the interface. This technology aims to accurately extract micro-slip information from complex contact signals. When micro-slip occurs between the gripper 4 and the target object, the tangential force on the interface generates high-frequency fluctuations. These fluctuations are characteristic signals of the transition from static friction to kinetic friction. By performing high-frequency filtering and time-domain integration on the tangential force signal detected by the tactile sensing array, the control unit can accumulate these transient fluctuations, thereby quantifying the transient micro-slip rate of the interface. For example, digital signal processors (DSPs) or field-programmable gate arrays (FPGAs) can be used to implement real-time acquisition and integration algorithms for high-frequency signals.

[0101] This processing method can effectively distinguish between macroscopic slip and microscopic slip, providing accurate input for the interface contact quality model, and extracting the maximum tensile principal strain value within the coverage area of ​​the strain sensor network as the maximum principal strain of the feature region. This technology is used to characterize the maximum local deformation of the gripper 4 during the gripping process. In the fluid-driven gripper 4, the deformation is complex and nonlinear. The maximum tensile principal strain reflects the maximum tensile deformation that the material can withstand in a certain direction, and is a key indicator for assessing the material failure risk and the gripper 4's encapsulation performance. The control unit analyzes the data from all sensors in the strain sensor network, uses the strain tensor analysis method to calculate the principal strain value at each measurement point, and selects the maximum tensile principal strain value within the network coverage area as the maximum principal strain of the characteristic region. For example, finite element analysis (FEA) or machine learning algorithms can be used to process sensor data to identify strain concentration areas and extract the maximum principal strain; or the principal strain values ​​of key areas can be directly calculated using a preset geometric model and sensor positions. This ensures an accurate grasp of the deformation state of the key areas of the gripper 4, providing precise deformation parameters for the fluid-structure interaction synergy model.

[0102] By installing a triaxial accelerometer at the base of the robotic arm, the vibration amplitude of the base can be monitored and acquired in real time. This data directly reflects the degree of dynamic interference from the external environment on the entire robotic arm system. Simultaneously, a piezoelectric accelerometer is integrated into the base of the gripper 4 to accurately detect the high-frequency vibration acceleration generated by the gripper 4 under active control. This enables the system to sense and quantify the dynamic response of the gripper 4 itself. A tactile sensor array and a strain sensor network are deployed on the surface of the gripper 4. The tactile sensor array not only provides the spatial distribution variance of the clamping contact force, but also... More importantly, the control unit can accurately calculate the transient micro-slip rate of the interface based on the high-frequency fluctuation characteristics of the tangential force it detects, using a time-domain integration algorithm. This solves the problem of traditional methods failing to capture micro-slippage, providing refined data for evaluating interface contact quality. The strain sensor network is responsible for monitoring the deformation of the gripper 4 surface. The control unit analyzes the network data and extracts the maximum tensile principal strain value within the coverage area as the maximum principal strain of the characteristic region. This directly reflects the local deformation state of gripper 4 under fluid drive and object contact. These are precise physical quantities obtained by the sensor hardware architecture and signal processing logic, such as the vibration amplitude of the base. Actively applied high-frequency vibration acceleration , Interface transient micro-slip rate and the maximum principal strain in the characteristic region These parameters are used as key input parameters and fed into the dynamic disturbance model, the interface contact quality model, and the fluid-structure interaction synergy model, respectively. For example, the amplitude of the base vibration. and actively applied high-frequency vibration acceleration It can be directly used for the calculation of dynamic disturbance models, enabling the system to accurately assess and compensate for external and internal dynamic disturbances.

[0103] Interface transient micro-slip rate This serves as a crucial component of the interface contact quality model, enabling the system to assess the contact stability between gripper 4 and the target object in real time. The maximum principal strain in the characteristic region... Together with the fluid medium characteristic coefficient, dynamic disturbance coefficient, and tip absolute displacement, a fluid-structure interaction synergy model is constructed to comprehensively evaluate the overall working state of gripper 4. This enables the control unit to output a more accurate target driving fluid absolute pressure based on a comprehensive evaluation of the fluid medium, dynamic disturbance, interface contact quality, and fluid-structure interaction synergy. This achieves high-precision, adaptive gripping control of the micro-manipulator gripping device. This combination of hardware deployment and signal processing effectively compensates for the shortcomings of traditional solutions in sensing micro-transient quantities, ensuring that the control model can obtain real and accurate system state information and avoiding control command lag or deviation caused by inaccurate data.

Claims

1. A fluid-driven micro manipulator gripping device, comprising an upper arm (1) and a forearm (2); Its features are: The forearm (2) is provided with a clamping part (3) and a control unit at its end. The clamping part (3) consists of two grippers (4), which are flexible. The control unit performs the following steps to achieve grasping control of the target object: Obtain the instantaneous temperature, instantaneous flow rate, and instantaneous ionic conductivity of the fluid, construct a fluid medium characteristic model, and output the fluid medium characteristic coefficients; The root mean square of the high-frequency pulsation of the driving pressure, the vibration amplitude of the base, and the actively applied high-frequency vibration acceleration are obtained to construct a dynamic disturbance model and output the dynamic disturbance coefficient. Obtain the spatial distribution variance of clamping contact force, the transient micro-slip rate of the interface, and the actual surface contact angle; construct an interface contact quality model and output the interface contact quality coefficient. By combining the fluid medium characteristic coefficient and dynamic disturbance coefficient, and obtaining the maximum principal strain and absolute displacement of the tip of the gripper (4), a fluid-structure interaction synergy model is constructed to output the synergy degree. Based on the degree of cooperation, the interface contact quality coefficient and the obtained current absolute pressure of the driving fluid, a pressure optimization model is constructed to output the target absolute pressure of the driving fluid to drive the control gripper (4).

2. The fluid-driven micro-manipulator gripping device according to claim 1, characterized in that: The fluid medium characteristic model calculates the dimensionless deviation between the instantaneous temperature, instantaneous flow rate, and instantaneous ionic conductivity of the fluid and their corresponding preset reference values, and combines them with their respective preset sensitivity weights. It adopts a negative correlation nonlinear decay mapping relationship to output fluid medium characteristic coefficients with values ​​greater than zero and less than or equal to one. The more the instantaneous parameters deviate from the reference values, the closer the fluid medium characteristic coefficients are to zero.

3. The fluid-driven micro-manipulator gripping device according to claim 2, characterized in that: The dynamic disturbance model described above divides the root mean square of the high-frequency pulsation of the driving pressure, the amplitude of the base vibration, and the actively applied high-frequency vibration acceleration by their respective preset safety thresholds for dimensionless processing. It then performs a weighted summation by combining preset weights for the influence of different disturbance sources. After deducting the preset system steady-state bias constant, it outputs the dynamic disturbance coefficient with a value range in the open interval from zero to one through an S-shaped nonlinear activation function with smooth saturation characteristics.

4. The fluid-driven micro-manipulator gripping device according to claim 3, characterized in that: The interface contact quality model extracts the spatial distribution variance of clamping contact force, the transient micro-slip rate of the interface, and the actual contact angle of the surface. It calculates the dimensionless relative deviations between these deviations and the preset non-uniformity tolerance variance threshold, the maximum allowable slip rate, and the optimal theoretical covering contact angle. It assigns a corresponding scaling factor to each deviation and performs a fusion mapping through a centrally symmetric bounded nonlinear translation function to output the interface contact quality coefficient with a value range in the open interval from zero to one.

5. A fluid-driven micro-manipulator gripping device according to claim 4, characterized in that: The fluid-structure interaction synergy model uses the ratio of the maximum principal strain in the characteristic region to the preset ultimate tensile strain rate, the fluid medium characteristic coefficient, the anti-interference margin after deducting the dynamic interference coefficient, and the penalty term for the absolute displacement of the tip deviating from the target optimal coverage expected displacement as independent dimensions. It performs global situation fusion through a multivariate nonlinear product structure based on preset elastic exponents of each dimension, and obtains the synergy in the range of zero to one, which is left-closed and right-open, by limiting the output boundary through an exponential saturation function.

6. The fluid-driven micro-manipulator gripping device according to claim 5, characterized in that: The pressure optimization model is based on the preset ratio of the desired synergy to the synergy, and the preset ratio of the desired contact quality to the interface contact quality coefficient. It calculates the corresponding feedback compensation terms through a logarithmic proportional function, and combines the preset gain coefficient to perform dynamic proportional adjustment calculations on the current driving fluid absolute pressure. The outermost layer of the pressure optimization model includes a cutoff protection mechanism based on physical extrema, ensuring that the absolute pressure of the output target driving fluid is always not lower than the preset safe pressure baseline.

7. A fluid-driven micro-manipulator gripping device according to claim 6, characterized in that: The fluid circuit of the driving gripper (4) integrates a temperature sensor for obtaining instantaneous temperature, a flow meter for obtaining instantaneous flow rate, and a conductivity probe for obtaining instantaneous ionic conductivity; the main channel of the fluid circuit is equipped with an absolute pressure sensor to detect the current absolute pressure of the driving fluid, and is equipped with a high-frequency dynamic piezoresistive pressure sensor to extract the fluid time domain signal and calculate the high-frequency pulsation root mean square of the driving pressure.

8. A fluid-driven micro-manipulator gripping device according to claim 7, characterized in that: A triaxial accelerometer is installed at the base of the robotic arm body to obtain the vibration amplitude of the base. A piezoelectric accelerometer is integrated at the bottom of the gripper (4) to detect the actively applied high-frequency vibration acceleration. A tactile sensing array and a strain sensor network are integrated on the surface of the gripper (4). The control unit performs time-domain integration based on the high-frequency fluctuation characteristics of the tangential force detected by the tactile sensing array to obtain the transient micro-slip rate of the interface, and extracts the maximum tensile principal strain value in the area covered by the strain sensor network as the maximum principal strain of the feature area.

Images