Magneto-rheological end effector accurate grasping system based on force feedback

Abstract

The invention relates to the technical field of robot intelligent actuators, and discloses a magneto-rheological end effector accurate grasping system based on force feedback, which comprises a multi-curved-surface bionic fingertip formed by a magneto-rheological elastomer, and a three-dimensional distributed excitation coil array is embedded in the fingertip. The coil array is excited by a high-frequency chopping constant-current driving module to generate a space gradient magnetic field, and the magnetic field changes the shear modulus of the elastomer within millisecond-level time, so that the fingertip surface hardness jumps in real time along with a control instruction; and the force feedback closed-loop control unit comprises an ultrathin film strain sensing net attached to the inner curved surface of the fingertip. Through time-sharing switching of a negative pressure micropore adsorption layer and a magneto-rheological rigidity keeping mode, traditional adsorption and rigid clamping are changed into a pre-fixing and rigidity gradual change two-section type process, instantaneous force impact during adsorption release is eliminated, meanwhile, thin film wrinkling caused by continuous negative pressure is avoided, and it is ensured that an ultrathin and high-added-value workpiece is kept flat and traceless in the whole transfer process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of household kitchen, in particular to a magnetorheological end effector precise grasping system based on force feedback. BACKGROUND

[0002] Robot end grasping technology has gone through four stages of rigid gripper, pneumatic suction cup, electrorheological damper and magnetorheological wrist joint, but still cannot meet the five demands of millisecond level stiffness transition, curved surface self-adaptation, zero impact, no continuous negative pressure and multi-dimensional force in-situ closed loop; rigid gripper realizes force-position hybrid control through servo motor combined with torque sensor, but the fingertip hardness is fixed, the inertia impact at the moment of contact is borne by the target, and micro cracks or permanent indentations appear on the glass substrate and flexible display film; pneumatic suction cup needs to quickly break the vacuum in the release stage, and the airflow reverse impact causes the center of the film to bulge and be accompanied by high-frequency vibration, and the sealing failure forms an adsorption blind area at the edge of the micro hole or curved surface, so the system has to increase the vacuum degree to compensate for the leakage, further aggravating the energy consumption and noise; the electrorheological damper relies on a kilovolt-level electric field, and the response speed is limited by the risk of breakdown, and the impedance drifts after the long-term operation of the carbonized insulation liquid, shortening the maintenance cycle; the magnetorheological wrist joint can generate stepless damping in the rotary degree of freedom, but the variable stiffness is limited to the driving end, and the fingertip contact interface is still a passive elastic pad, and the force control closed loop needs to pass through three levels of motor reducer, connecting rod and flexible pad, the phase lag is more than ten milliseconds, and the suppression ability to slip or external impact is greatly weakened; on the sensing level, the traditional scheme arranges single-axis strain gauges at the finger root or vacuum pressure switches at the edge of the suction cup, only discrete point information can be obtained, the shear direction angle and the full-field strain distribution cannot be obtained, the controller cannot distinguish between target active slip and robot active detection, and excessive clamping force is triggered after misjudgment, forming positive feedback damage; scenes such as consumer electronics, new energy battery films and biological biopsy tissues require that the grasping process be vision-independent, surface-defect-free and airflow-noise-free.

[0003] The existing technology has the problems of long stiffness adjustment link, inability to immediately remove negative pressure and lack of multi-dimensional force information, resulting in reliability and yield bottlenecks in front of ultra-thin, fragile, curved and eccentric mass targets, and there is an urgent need for a complete system that directly arranges magnetorheological elastomers at the fingertips, integrates a three-dimensional distributed excitation array, a multi-dimensional force sensing network and an event-triggered protection mechanism, to realize sub-millisecond hardness transition, negative pressure pre-fixing and seamless switching of stiffness, rapid dissipation of impact energy and vision-free zero-defect grasping. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a magnetorheological end effector precise grasping system based on force feedback, which solves the problems raised in the background art.

[0005] In order to achieve the above object, the application is realized by the following technical solutions: a precise gripping system of a magneto-rheological end effector based on force feedback, comprising: A multi-curved surface bionic fingertip composed of a magneto-rheological elastomer, a three-dimensional distributed excitation coil array is embedded in the inside of the fingertip, the coil array generates a spatial gradient magnetic field after being excited by a high-frequency chopper constant current driving module, the magnetic field changes the shear modulus of the elastomer within milliseconds, and the surface hardness of the fingertip jumps with the real-time control instruction; A force feedback closed-loop control unit, the force feedback closed-loop control unit includes an ultrathin film strain sensing network attached to the inner curved surface of the fingertip, a micro-nano piezoresistive sensing network attached to the outer curved surface of the fingertip, and a shear force fiber grid buried in the inside of the elastomer, the sensing network and the grid output multi-dimensional force signals synchronously, the signals generate hardness correction instructions after being operated by a hardware-in-the-loop programmable gate array, and the instructions are sent back to the excitation coil array to form a hardness-force sensing two-way closed loop; A multi-joint compliant wrist, the multi-joint compliant wrist is built-in a magneto-rheological damping hub, the hub generates variable impedance according to the damping adjustment instruction output by the force feedback closed-loop control unit during the grabbing process, so that the end effector presents an inertia offset state at the moment of contacting the target; A negative pressure micropore adsorption layer, the negative pressure micropore adsorption layer is covered on the surface of the fingertip and conforms to the magneto-rheological elastomer, the layer provides instantaneous negative pressure pre-fixing at the initial stage of gripping, and actively closes the negative pressure and switches to a pure magneto-rheological stiffness maintaining mode after the gripping is stable; A central multi-core real-time processor, the central multi-core real-time processor communicates synchronously with the high-frequency chopper constant current driving module, the hardware-in-the-loop programmable gate array, and the multi-joint compliant wrist through time-triggered Ethernet, the processor is built-in a nonlinear inverse model, the inverse model takes force signals as input and spatial gradient magnetic field strength distribution as output, and realizes sub-millisecond predictive compensation of the gripping force trajectory.

[0006] Preferably, the three-dimensional distributed excitation coil array adopts an asymmetric saddle-shaped winding layout, high-magnetic-permeability nanocrystalline spacers are embedded between adjacent windings, the spacers keep the residual magnetic flux below a preset threshold after power-off, so as to avoid magnetic hysteresis adhesion during grabbing and releasing.

[0007] Preferably, the ultrathin film strain sensing network is formed by laser micro-etching a two-dimensional transition metal carbide film into a serpentine conductive grid, the grid generates a tunnel resistance change when the elastomer deforms, the change is output after carrier modulation and demodulation to output a micro-strain field distribution map, and the distribution map is used to correct the boundary conditions of the inverse model.

[0008] Preferably, the shear force fiber grating is installed in a spiral manner, and an ultra-short Bragg grating string is written into the fiber core. When the ultra-short Bragg grating string is sheared, it generates a chirped reflection spectrum. The spectrum is extracted by matched filtering to obtain the shear force direction angle, which is used to distinguish between the target sliding trend and the active detection action.

[0009] Preferably, the magnetorheological damping hub of the multi-joint compliant wrist includes a dual-disc rotor and a stator. The rotor surface is covered with a wear-resistant ceramic coating, and the inner ring of the stator is provided with a cooling spiral microchannel. The channel is filled with an insulating magnetorheological fluid. During the damping adjustment process, the insulating magnetorheological fluid simultaneously removes the Joule heat from the excitation coil, thereby achieving thermal-mechanical coupling stability.

[0010] Preferably, the micropores of the negative pressure microporous adsorption layer have a biomimetic tree-like fractal structure, and the micropore outlet diameter gradually decreases along the layer thickness direction. After the negative pressure is closed, the structure retains the air pad through capillary force, so that the target surface is protected from rigid impact during the stage of increasing magnetorheological stiffness.

[0011] Preferably, the hardware deploys a random spiking neural network inside a ring-programmable gate array. The network takes a force signal pulse sequence as input and a hardness-corrected pulse width modulation code as output. The network weights are updated online during the grasping process by an on-site reinforcement learning engine. The reward function of the engine is negatively correlated with the target sliding amount.

[0012] Preferably, the central multi-core real-time processor reserves an event trigger interface. When an external impact disturbance is detected, the interface interrupts the normal gripping program and starts the magnetorheological elastomer instantaneous high stiffness mode. The duration of the mode is dynamically determined by the disturbance energy estimation module, which calculates the energy integral based on the disturbance acceleration amplitude and duration.

[0013] Preferably, the high-frequency chopper constant current drive module adopts a three-phase four-quadrant H-bridge cascaded topology. The switching frequency of the topology is adaptively adjusted according to the rate of change of the spatial gradient magnetic field output by the inverse model. The adjustment reduces electromagnetic noise through a sliding mode harmonic suppression algorithm, so that the system meets the requirements of silent operation scenarios.

[0014] Preferably, when the magnetorheological end effector grasps a fragile thin film target, the negative pressure microporous adsorption layer first provides stress-free pre-fixation, and then the force feedback closed-loop control unit gradually reduces the current of the excitation coil array, so that the magnetorheological elastomer undergoes a three-stage soft landing of high stiffness, medium stiffness and low stiffness. During the soft landing process, the shear force fiber optic grid continuously monitors the target wrinkle wavelength. Once the wavelength exceeds the preset wrinkle threshold, the central multi-core timing processor immediately increases the stiffness to suppress wrinkle expansion, thereby achieving a visually independent and smooth film grasp.

[0015] This invention provides a precision gripping system for a magnetorheological end effector based on force feedback. It offers the following advantages: 1. This invention uses magnetorheological elastomers, three-dimensional distributed excitation coil arrays, multi-dimensional force sensing network hardness and force sensing bidirectional closed loop to compress the fingertip hardness transition time to the millisecond level, so that the gripping force can complete adaptive matching at the moment of contact, significantly reducing the risk of target slippage and overpressure damage, and realizing visual precision control of fragile targets.

[0016] 2. This invention uses a time-sharing switching between the negative pressure microporous adsorption layer and the magnetorheological stiffness holding mode, transforming the traditional adsorption and rigid clamping into a two-stage process of pre-fixation and gradual stiffness change. This eliminates the instantaneous force impact during adsorption release and avoids film wrinkling caused by continuous negative pressure, ensuring that ultra-thin, high-value-added workpieces remain flat and traceless throughout the transfer process.

[0017] 3. The nonlinear inverse model and event-triggered high-rigidity protection mechanism built into the central multi-core real-time processor of this invention work together to provide sub-millisecond-level predictive compensation for grip force trajectory under normal conditions, and to instantly improve fingertip stiffness under sudden impact, thus balancing the compliance of daily operation with the rigid protection against unexpected disturbances, and greatly improving the reliability and lifespan of the system under dynamic operating conditions. Attached Figure Description

[0018] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see the appendix Figure 1 This invention provides a force feedback-based magnetorheological end effector precision gripping system, comprising: The multi-curved biomimetic fingertip is made of magnetorheological elastomer. The fingertip has a three-dimensional distributed excitation coil array embedded inside. After the coil array is excited by the high-frequency chopper constant current drive module, it generates a spatial gradient magnetic field. The magnetic field changes the shear modulus of the elastomer in milliseconds, so that the surface hardness of the fingertip jumps in real time with the control command. The three-dimensional distributed excitation coil array adopts an asymmetric saddle-shaped winding layout. High permeability nanocrystalline spacers are embedded between adjacent windings. After the power is cut off, the spacers keep the residual magnetic flux below a preset threshold to avoid magnetic hysteresis adhesion during the grasping and releasing phase.

[0021] Specifically, the magnetorheological elastomer exhibits a low shear modulus and macroscopic softness in a zero-field state. When a high-frequency chopped constant current is applied to the three-dimensional distributed excitation coil array, the asymmetric saddle-shaped winding superimposes a spatial gradient magnetic field inside the fingertip. The direction and magnitude of the magnetic field change with position, causing the magnetic particles inside the elastomer to form a chain-like structure along the magnetic field lines. The interaction between the particles instantly increases the macroscopic shear modulus, and the surface hardness of the fingertip jumps accordingly. After the power is turned off, the high-permeability nanocrystalline spacer provides a low-resistance magnetic circuit, which quickly guides the remaining magnetic flux into the spacer, causing the field strength in the elastomer region to drop rapidly below the threshold. The chain-like structure disintegrates, and the hardness returns to softness, thus completing a stiffness switching cycle without magnetic adhesion.

[0022] The force feedback closed-loop control unit includes an ultra-thin film strain sensor network attached to the inner curved surface of the fingertip, a micro-nano piezoresistive sensor network attached to the outer curved surface of the fingertip, and a shear force fiber optic grid embedded inside the elastic body. The sensor network and the grid synchronously output multi-dimensional force signals. After the signals are processed by the hardware-in-loop programmable gate array, a hardness correction command is generated. The command is sent back to the excitation coil array to form a hardness-force bidirectional closed loop. The ultra-thin film strain sensor network is formed by laser micro-etching of a two-dimensional transition metal carbide film into a serpentine conductive grid. When the elastic body deforms, the grid generates a tunnel resistance change. After carrier modulation and demodulation, the change outputs a micro-strain field distribution map, which is used to correct the boundary conditions of the inverse model.

[0023] Specifically, when the target comes into contact with the fingertip, the serpentine two-dimensional transition metal carbide mesh of the inner curved ultra-thin film strain sensor network changes its tunnel resistance as the elastic body is stretched. The outer curved micro-nano piezoresistive sensor network records the contact pressure, and the embedded fiber optic grid senses the shear direction. The three signals are simultaneously sent to a hardware-in-the-loop programmable gate array. The array uses an embedded random pulse neural network to fuse multi-dimensional force information in real time and outputs a hardness correction pulse width modulation code to drive the excitation coil array to adjust the spatial gradient magnetic field, thereby changing the shear modulus of the elastic body and increasing or decreasing the fingertip stiffness. This, in turn, changes the sensing signal, forming a closed loop. The full-field micro-strain distribution map provided by the serpentine mesh is used to update the boundary conditions of the inverse model online, compensate for the errors caused by temperature and creep, and ensure that the closed-loop bandwidth is always in the sub-millisecond range, achieving a dynamic balance between force sensing and hardness.

[0024] The shear force fiber grating is installed in a spiral manner, with an ultra-short Bragg grating string written into the fiber core. When the ultra-short Bragg grating string is sheared, it generates a chirped reflection spectrum. The spectrum is extracted by matched filtering to obtain the shear force direction angle, which is used to distinguish between the target sliding trend and the active detection action.

[0025] Specifically, a shear force fiber grating is embedded in the magnetorheological elastomer via a helical path. When the elastomer is subjected to tangential force and undergoes microscale shear deformation, the period of the ultrashort Bragg grating string within the helical segment undergoes non-uniform expansion and contraction along the force direction, resulting in linear chirped broadening of the reflection spectrum. The broadening slope is proportional to the shear magnitude, and the direction of the broadening dominant frequency shift corresponds to the orientation of the tangential vector. A hardware-in-the-loop programmable gate array performs matched filtering on the chirped spectrum to calculate the shear force direction angle. If the direction angle is consistent with the preset motion vector actively detected by the fingertip, it is determined to be the robot's own action, and stiffness enhancement is not triggered. If the direction angle deviates from the preset vector and the amplitude continues to increase, it is determined to be a target sliding trend, and a stiffness correction command is immediately output to drive the excitation coil array to increase the stiffness of the elastomer, thereby achieving rapid differentiation between sliding warning and active suppression under visual conditions.

[0026] The multi-joint compliant wrist incorporates a magnetorheological damping hub. During the grasping process, the hub generates variable impedance based on the damping adjustment command output by the force feedback closed-loop control unit, causing the end effector to exhibit inertia cancellation at the moment of contact with the target. The magnetorheological damping hub of the multi-joint compliant wrist includes a dual-disc rotor and stator. The rotor surface is covered with a wear-resistant ceramic coating, and the inner ring of the stator has a cooling spiral microchannel. Insulating magnetorheological fluid circulates in the channel. During the damping adjustment process, the insulating magnetorheological fluid simultaneously removes Joule heat from the excitation coil, achieving thermal-mechanical coupling stability.

[0027] Specifically, the magnetorheological damping hub of the multi-joint compliant wrist adopts a double-disc structure. The gap between the rotor and stator is filled with insulating magnetorheological fluid. When the force feedback closed-loop control unit issues a damping adjustment command, the excitation coil establishes a radial magnetic field in the gap, and the magnetorheological fluid instantaneously forms a chain structure. The shear yield strength changes continuously with the field strength, and the hub impedance can rise from zero damping to high damping in milliseconds, so that the end inertia is actively absorbed at the moment of contact with the target, achieving a soft landing without rebound. The ceramic coating on the rotor surface reduces high-speed friction and wear, and the spiral microchannel in the stator drives the insulating magnetorheological fluid to circulate, disengaging the Joule heat of the coil from the gap, keeping the liquid phase temperature constant, and avoiding thermal viscosity drift. Thus, damping adjustment and heat dissipation are coupled in the same liquid path, ensuring dual thermal and mechanical stability under long-term variable impedance conditions.

[0028] The negative pressure microporous adsorption layer covers the fingertip surface and conforms to the magnetorheological elastomer. The layer provides instantaneous negative pressure pre-fixation at the beginning of the gripping stage. After the grip stabilizes, the force feedback closed-loop control unit actively closes the negative pressure and switches to a pure magnetorheological stiffness maintenance mode. The micropores of the negative pressure microporous adsorption layer have a biomimetic tree-like fractal structure. The diameter of the micropore outlet gradually decreases along the thickness direction of the layer. After the negative pressure is closed, the structure retains the air pad through capillary force, so that the target surface is protected from rigid impact during the magnetorheological stiffness increase stage.

[0029] Specifically, the negative pressure microporous adsorption layer conformally adheres to the surface of the magnetorheological elastomer. Its internal biomimetic tree-like fractal channels connect to the vacuum source at the moment of gripping initiation. The micropores at the exit end, which gradually narrow, generate a high-speed, low-pressure zone, initially pulling the target gently towards the fingertip in a stress-free manner to complete pre-fixation. When the force feedback closed-loop control unit detects that the multi-dimensional force signal tends to stabilize, it immediately shuts off the vacuum and turns on the magnetorheological excitation. At this time, the small amount of air remaining in the fractal channel is locked in the progressively narrowing pores due to capillary force, forming a uniformly distributed micron-level air cushion. This air cushion acts as a flexible buffer during the transition phase when the shear modulus of the elastomer rapidly increases, significantly smoothing the rate of change of contact stiffness on the target surface. Subsequently, the air cushion gas slowly diffuses out, and the fingertip completely enters the high-stiffness holding mode, achieving a seamless connection between negative pressure release and rigid clamping, avoiding the reverse impact and local indentation during the release of pressure from traditional suction cups.

[0030] The central multi-core real-time processor communicates synchronously with the high-frequency chopper constant current drive module, hardware-in-the-loop programmable gate array, and multi-joint compliant wrist via time-triggered Ethernet. The processor has a built-in nonlinear inverse model, which takes the force signal as input and the spatial gradient magnetic field intensity distribution as output to achieve sub-millisecond predictive compensation of the gripping force trajectory.

[0031] Specifically, the central multi-core time processor uses time-triggered Ethernet clock cycles to unify the high-frequency chopper constant current drive module, hardware-in-the-loop programmable gate array, and clock reference of the multi-joint flexible wrist, eliminating communication jitter. Its built-in nonlinear inverse model maps the ultra-thin film strain, micro-nano piezoresistive force, and fiber shear force signals from the fingertips into a spatial gradient magnetic field intensity distribution. Through feedforward calculation, it adjusts the excitation current of each voxel in advance before force deviation occurs, so that the shear modulus of the magnetorheological elastomer jumps synchronously with the target morphology and load changes, achieving sub-millisecond predictive compensation of the gripping force trajectory. This ensures that the contact force always tracks the ideal curve without waiting for error feedback, thereby compressing the closed-loop phase lag to the minimum and supporting zero overshoot stable gripping under high dynamic conditions.

[0032] The hardware deploys a random spiking neural network inside a ring-programmable gate array. The network takes a force signal pulse sequence as input and a hardness-corrected pulse width modulation code as output. The network weights are updated online during the grasping process by an on-site reinforcement learning engine. The engine's reward function is negatively correlated with the target sliding amount.

[0033] Specifically, a hardware-in-the-loop programmable gate array encodes multi-dimensional force signals from ultra-thin film strain, micro-nano piezoresistive data, and fiber optic gratings into sparse pulse sequences, which are then input into a random spiking neural network. The on-site reinforcement learning engine uses negative target slip as an immediate reward and employs local plasticity rules to update synaptic weights online. The more pronounced the slipping trend, the lower the reward. Weight adjustments rapidly increase the network output duty cycle, thereby increasing the stiffness of the magnetorheological elastomer, suppressing the rebound of the reward after slipping, and causing the weights to converge. This achieves adaptive stiffness strategy generation for unknown object mass, friction, and deformation during the grasping process, without requiring offline training or external labels.

[0034] The central multi-core real-time processor reserves an event trigger interface. When an external impact disturbance is detected, the interface interrupts the normal grasping program and initiates the magnetorheological elastomer instantaneous high-stiffness mode. The duration of the mode is dynamically determined by the disturbance energy estimation module, which calculates the energy integral based on the disturbance acceleration amplitude and duration. The high-frequency chopper constant current drive module adopts a three-phase four-quadrant H-bridge cascaded topology. The switching frequency of the topology is adaptively adjusted according to the rate of change of the spatial gradient magnetic field output by the inverse model. The adjustment reduces electromagnetic noise through a sliding mode harmonic suppression algorithm, enabling the system to meet the requirements of silent operation scenarios.

[0035] Specifically, the event triggering interface of the central multi-core real-time processor continuously monitors the acceleration amplitude and duration. Once the integral value exceeds the set threshold, the normal gripping process is immediately interrupted, and a high-stiffness command is injected into the inverse model. At the same time, the disturbance energy estimation module outputs a duration counter based on the integral result. When the timing ends, it automatically returns to normal, realizing instantaneous hardening and timed rebound against external impacts. During this process, the three-phase four-quadrant H-bridge cascaded topology of the high-frequency chopper constant current drive module reads the spatial gradient magnetic field change rate given by the inverse model. The change rate is used as input to adjust the switching frequency in real time. When the frequency increases, the sliding mode harmonic suppression algorithm dynamically switches the sliding surface, causing the current harmonic energy to migrate to the high-frequency band and be efficiently attenuated by the LC filter. The electromagnetic noise power spectral density decreases, ensuring that the system maintains low-noise operation in quiet laboratories or nighttime work environments.

[0036] When the magnetorheological end effector grasps a fragile thin film target, it first provides stress-free pre-fixation through a negative pressure microporous adsorption layer, and then the force feedback closed-loop control unit gradually reduces the current of the excitation coil array, so that the magnetorheological elastomer undergoes a three-stage soft landing of high stiffness, medium stiffness and low stiffness. During the soft landing process, the shear force fiber optic grid continuously monitors the target wrinkling wavelength. Once the wavelength exceeds the preset wrinkling threshold, the central multi-point timing processor immediately increases the stiffness to suppress wrinkle expansion, thus achieving a smooth film grasp without visual dependence.

[0037] Specifically, during the gripping initiation, the negative pressure microporous adsorption layer gently adheres the film with uniform low pressure, avoiding initial stress concentration. Subsequently, the force feedback closed-loop control unit reduces the excitation current in a decreasing sequence, and the magnetorheological elastomer transitions sequentially from high stiffness to medium stiffness and then to low stiffness. The elastic modulus gradient of the fingertip contact surface decreases, and the contact force is dispersed over a larger area, achieving a three-stage soft landing. During this process, the spirally embedded shear force fiber optic grid converts the micro-shear caused by film wrinkling into a chirped reflection spectrum in real time. The central multi-core real-time processor extracts the wrinkle wavelength and compares it with a set threshold. If the wavelength exceeds the limit, the processor immediately reverses and increases the excitation current, causing a sharp increase in the elastomer's shear modulus. The contact surface hardens rapidly, inhibiting further wrinkle expansion. The entire process achieves a smooth and damage-free gripping of the film without visual feedback.

[0038] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A precision gripping system for a magnetorheological end effector based on force feedback, characterized in that, include: The multi-curved biomimetic fingertip is made of magnetorheological elastomer. The fingertip has a three-dimensional distributed excitation coil array embedded inside. After the coil array is excited by a high-frequency chopper constant current drive module, it generates a spatial gradient magnetic field. The magnetic field changes the shear modulus of the elastomer in milliseconds, so that the surface hardness of the fingertip changes in real time with the control command. The force feedback closed-loop control unit includes an ultra-thin film strain sensor network attached to the inner curved surface of the fingertip, a micro-nano piezoresistive sensor network attached to the outer curved surface of the fingertip, and a shear force fiber optic grid embedded in the elastic body. The sensor network and the grid synchronously output multi-dimensional force signals. The signals are processed by a hardware-in-loop programmable gate array to generate a hardness correction command. The command is sent back to the excitation coil array to form a hardness-force bidirectional closed loop. The multi-joint compliant wrist has a built-in magnetorheological damping hub. During the grasping process, the hub generates variable impedance according to the damping adjustment command output by the force feedback closed-loop control unit, so that the end effector presents an inertia cancellation state at the moment of contact with the target. A negative pressure microporous adsorption layer covers the fingertip surface and conforms to the magnetorheological elastomer. The layer provides instantaneous negative pressure pre-fixation during the initial stage of gripping. After the gripping is stable, the force feedback closed-loop control unit actively shuts off the negative pressure and switches to a pure magnetorheological stiffness retention mode. The central multi-core time processor communicates synchronously with the high-frequency chopper constant current drive module, the hardware-in-the-loop programmable gate array, and the multi-joint compliant wrist via time-triggered Ethernet. The processor has a built-in nonlinear inverse model, which takes the force signal as input and the spatial gradient magnetic field intensity distribution as output to achieve sub-millisecond predictive compensation of the gripping force trajectory.

2. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The three-dimensional distributed excitation coil array adopts an asymmetric saddle-shaped winding layout, with high permeability nanocrystalline spacers embedded between adjacent windings. After power is cut off, the spacers maintain the residual magnetic flux below a preset threshold to avoid hysteresis adhesion during the grasping and releasing phase.

3. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The ultrathin film strain sensing network is formed by laser micro-etching of a two-dimensional transition metal carbide film into a serpentine conductive grid. The grid generates a tunnel resistance change when the elastic body deforms. The change is modulated by a carrier wave and then output as a micro-strain field distribution map. The distribution map is used to correct the boundary conditions of the inverse model.

4. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The shear force fiber grating is installed in a spiral manner, and an ultra-short Bragg grating string is written into the fiber core. When the ultra-short Bragg grating string is sheared, it generates a chirped reflection spectrum. The spectrum is extracted by matched filtering to obtain the shear force direction angle, which is used to distinguish between the target sliding trend and the active detection action.

5. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The magnetorheological damping hub of the multi-joint compliant wrist includes a dual-disc rotor and a stator. The rotor surface is covered with a wear-resistant ceramic coating, and the inner ring of the stator is provided with a cooling spiral microchannel. The channel is filled with an insulating magnetorheological fluid. During the damping adjustment process, the insulating magnetorheological fluid simultaneously removes the Joule heat from the excitation coil, thereby achieving thermal-mechanical coupling stability.

6. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The micropores of the negative pressure microporous adsorption layer have a biomimetic tree-like fractal structure. The diameter of the micropore outlet gradually decreases along the thickness direction. After the negative pressure is closed, the structure retains the air pad through capillary force, so that the target surface is protected from rigid impact during the stage of increasing magnetorheological stiffness.

7. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The hardware deploys a random spiking neural network inside a ring-programmable gate array. The network takes a force signal pulse sequence as input and a hardness-corrected pulse width modulation code as output. The network weights are updated online during the grasping process by an on-site reinforcement learning engine. The reward function of the engine is negatively correlated with the target sliding amount.

8. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The central multi-core real-time processor has a reserved event trigger interface. When an external impact disturbance is detected, the interface interrupts the normal gripping program and starts the magnetorheological elastomer instantaneous high stiffness mode. The duration of the mode is dynamically determined by the disturbance energy estimation module, which calculates the energy integral based on the disturbance acceleration amplitude and duration.

9. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, The high-frequency chopper constant current drive module adopts a three-phase four-quadrant H-bridge cascaded topology. The switching frequency of the topology is adaptively adjusted according to the rate of change of the spatial gradient magnetic field output by the inverse model. The adjustment reduces electromagnetic noise through a sliding mode harmonic suppression algorithm, so that the system meets the requirements of silent operation scenarios.

10. The magnetorheological end effector precision gripping system based on force feedback according to claim 1, characterized in that, When the magnetorheological end effector grasps a fragile thin film target, it first provides stress-free pre-fixation through the negative pressure microporous adsorption layer, and then the force feedback closed-loop control unit gradually reduces the current of the excitation coil array, so that the magnetorheological elastomer undergoes a three-stage soft landing of high stiffness, medium stiffness and low stiffness. During the soft landing process, the shear force fiber optic grid continuously monitors the target wrinkling wavelength. Once the wavelength exceeds the preset wrinkling threshold, the central multi-point timing processor immediately increases the stiffness to suppress wrinkle expansion, thus achieving a visually independent and smooth film grasp.

Images