Bionic flexible self-adaptive dexterous hand and control method

Abstract

The invention relates to the technical field of robot end executors, in particular to a bionic flexible self-adaptive dexterous hand and a control method.The dexterous hand comprises a palm, bionic fingers, a driving assembly, a reset assembly and a sensing system, the bionic fingers comprise knuckles connected in series through joint mechanisms, and the bionic fingers are movably connected with the palm through finger root joints; a non-Newtonian fluid core-shell structure is arranged on the finger pulp side and comprises a flexible sealing shell and a non-Newtonian fluid inner core filled in the flexible sealing shell, the driving assembly drives the fingers to bend through tendon rope retracting and releasing, the reset assembly enables the fingers to recover stretching after driving force is relieved, and the sensing system integrates a flexible sensor to feed back the grabbing state in real time. The non-Newtonian fluid slowly flows under pressure, so that the flexible shell perfectly fits the outline of an object, the contact area and friction force are greatly increased, creep deformation is resisted due to high internal energy consumption, and the irregular and smooth object is stably grabbed.

Description

Technical Field

[0001] This invention relates to the field of robot end effector technology, specifically to a biomimetic flexible adaptive dexterous hand and its control method. Background Technology

[0002] With the continuous development of robotics technology, bionic dexterous hands are increasingly widely used in many fields such as industrial automated production, space exploration, agricultural harvesting, and service robots. Unlike traditional rigid grippers, bionic dexterous hands strive to achieve good compliance and adaptability while ensuring mechanical control precision, so as to achieve safe and reliable grasping in complex environments.

[0003] However, most existing robotic dexterous hands use rigid materials or solid elastomers (such as silicone) for their contact surfaces. While rigid contact surfaces have high load-bearing capacity, they have poor adaptability when facing irregular or smooth objects and are prone to slippage or damage. Solid elastomers, on the other hand, have a certain degree of flexibility, but their deformation and recovery depend entirely on the elasticity of the material itself. Their damping characteristics are limited, and they cannot effectively absorb impacts or suppress vibrations. At the same time, the hardness of these materials is fixed and difficult to adjust according to task requirements. This results in a lack of a controllable transition state between soft and hard states during the grasping process, which affects the stability and reliability in complex scenarios.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a biomimetic flexible adaptive dexterous hand and control method, thereby effectively solving the problems pointed out in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A biomimetic flexible adaptive dexterous hand includes: a palm, multiple biomimetic fingers, an actuation component, a reset component, and a sensing system; The bionic finger includes multiple phalanges connected in series by a joint mechanism, and the bionic finger is movably connected to the palm through the phalangeal joint; At least a portion of the finger joints have a non-Newtonian fluid core-shell structure on the fingertip side, the non-Newtonian fluid core-shell structure comprising a closed cavity formed by a flexible sealing shell and a non-Newtonian fluid core filling the closed cavity. The drive assembly includes a tendon cord and a drive component that pass through the internal channel of the knuckle, and drives the bending motion of the bionic finger by extending and retracting the tendon cord. The reset assembly includes an elastic reset mechanism disposed on the back side of the knuckle, which is used to restore the knuckle to an extended state after the driving force of the driving assembly is released. The sensing system includes a flexible sensor integrated on or inside the non-Newtonian fluid core-shell structure for real-time detection and feedback of the detection and grasping status information.

[0007] Furthermore, the flexible sealing shell of the non-Newtonian fluid core-shell structure is a multi-layer composite structure, including at least an inner sealing layer, an intermediate reinforcing braided layer, and an outer wear-resistant layer.

[0008] Furthermore, the non-Newtonian fluid core is at least one of a shear-thickening fluid, a magnetorheological fluid, or an electrorheological fluid.

[0009] Furthermore, when the non-Newtonian fluid core is a magnetorheological fluid or an electrorheological fluid, a fluid state control system is also included, which includes a control device for applying a magnetic field or an electric field to the non-Newtonian fluid core-shell structure.

[0010] The present invention also includes a control method for a biomimetic flexible adaptive dexterous hand as described above, the method comprising: Adaptive envelope stage: Control the dexterous hand to make the gripping surface with the non-Newtonian fluid core-shell structure contact the object at a predetermined contact speed, and maintain the non-Newtonian fluid core-shell structure in a low viscosity state, so that the gripping surface produces flexible deformation to passively fit and enclose the object contour. Rigidification and locking phase: When the preset triggering conditions are met, the non-Newtonian fluid core-shell structure is controlled to switch from the low viscosity state to the high viscosity state, thereby increasing its apparent viscosity and making the gripping surface rigid to lock the gripping posture.

[0011] Furthermore, the non-Newtonian fluid core-shell structure can be controlled to switch to a high viscosity state in any of the following ways: When the non-Newtonian fluid core is a shear-thickening fluid, a high viscosity state is triggered by the slight tightening action of the drive component; When the non-Newtonian fluid core is a magnetorheological fluid or an electrorheological fluid, high viscosity switching is achieved by applying an external magnetic field or electric field through a fluid state control system.

[0012] Furthermore, the triggering condition is determined through the following steps: Real-time acquisition of multimodal sensing data provided by the sensing system, including pressure distribution, strain distribution and vibration signals; Extract temporal features from the multimodal sensing data; The temporal features are input into a pre-trained stability evaluation model to obtain the slip probability of the current grasping state; When the slippage probability exceeds the dynamic threshold, the triggering condition is determined to be met.

[0013] Further, the temporal features are input into a pre-trained stability evaluation model to obtain the slip probability of the current grasping state. The steps include: The multimodal sensing data is segmented according to a preset time window, and the pressure, strain, and vibration signals are synchronized in time. The contact area change rate, normal force fluctuation, strain gradient and spectral energy characteristics are extracted from the time synchronization data, and the characteristics are normalized. The processed features are input into the temporal coding module to generate temporal embedding vectors; The temporal embedding vector is input into a pre-trained stability evaluation model, which outputs the slip probability of the current grasping state. The slip probability is smoothed and compared with a dynamic threshold that is adaptively updated based on the normal force and friction coefficient. When the slip probability exceeds the threshold continuously, a trigger signal is output.

[0014] Furthermore, after completing the rigidification locking and entering the object transport process, it also includes: Dynamic stabilization phase: During the transport of the object, the dynamic signals of vibration acceleration or pressure distribution fluctuations are monitored in real time through the sensing system. Based on the dynamic signal, the viscosity of the non-Newtonian fluid core-shell structure is dynamically adjusted in real time to actively suppress slippage or vibration.

[0015] Furthermore, the real-time dynamic adjustment is achieved by applying a pulsating control field, and the frequency of the pulsation is related to the dominant frequency of the dynamic signal and controlled in an anti-phase manner to counteract vibration.

[0016] The technical solution of this invention can achieve the following technical effects: 1. Excellent adaptive envelope capability: The non-Newtonian fluid flows slowly under pressure, allowing the flexible shell to perfectly conform to the contour of the object, greatly increasing the contact area and friction. Furthermore, due to its high internal energy dissipation, it resists creep and achieves stable gripping of irregular and smooth objects. 2. Excellent damping and vibration reduction: The high damping characteristics of non-Newtonian fluids can effectively dissipate impact kinetic energy, significantly reduce impact vibration and rebound during the grasping process, and protect fragile items and hand structures. 3. Actively adjustable mechanical properties: By selecting the fluid type, adjusting the pressure or introducing an external field, the hardness, viscosity and damping characteristics of the contact surface can be changed in real time and reversibly, realizing the intelligent switching of "soft envelopment during grasping and rigid support during operation", adapting to various task requirements such as precision picking and stable gripping. 4. High functional integration: The fluid itself can serve as a sensing medium. By monitoring changes in its pressure or flow field, high-precision tactile perception can be achieved without additional contact sensors, simplifying the structure. 5. The structural advantages of combining rigidity and flexibility: The internal rigid frame ensures motion accuracy and load capacity, while the external fluid contact surface provides intelligent compliance. Combined with multi-degree-of-freedom drive, it achieves a balance of flexibility, stability and safety. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram of the structure of a biomimetic flexible adaptive dexterous hand; Figure 2 This is a schematic diagram of the structure of the lateral fingers; Figure 3 A schematic diagram of the structure of other fingers; Figure 4 This is a schematic diagram of the elastic reset mechanism; Figure 5 This is a schematic diagram of the joint mechanism; Figure 6 This is a schematic diagram of a non-Newtonian fluid core-shell structure. Figure 7 A flowchart illustrating the control method for a biomimetic flexible adaptive dexterous hand; Figure 8 A flowchart for determining the triggering conditions.

[0019] Reference numerals: 1. Palm; 2. Finger; 21. Knuckle; 22. Joint mechanism; 221. Sliding bearing; 222. Interdigital rivet; 23. Finger root joint; 231. Palm knuckle; 232. Palm sliding bearing; 3. Non-Newtonian fluid core-shell structure; 31. Flexible sealed outer shell structure; 32. Non-Newtonian fluid inner core; 33. Concave-convex structure; 4. Thigh ligament; 5. Elastic reset mechanism; 6. Flexible sensor. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] like Figures 1 to 6 As shown: A biomimetic flexible adaptive dexterous hand includes: a palm 1, multiple biomimetic fingers 2, a driving component, a reset component, and a sensing system; The bionic finger 2 includes multiple phalanges 21 connected in series by joint mechanisms 22. The bionic finger 2 is movably connected to the palm 1 through the root joint 23. At least a portion of the finger joint 21 has a non-Newtonian fluid core-shell structure 3 on the fingertip side. The non-Newtonian fluid core-shell structure 3 includes a closed cavity formed by a flexible sealed outer shell 31 and a non-Newtonian fluid core 32 filled in the closed cavity. The drive assembly includes a tendon cord 4 that passes through the internal channel of the knuckle 21 and a drive component, which drives the bending motion of the bionic finger 2 by retracting and extending the tendon cord 4. The reset assembly includes an elastic reset mechanism 5 disposed on the back side of the knuckle 21, which is used to restore the knuckle 21 to the extended state after the driving force of the drive assembly is released; specifically, a pre-compressed spring or an elastic sheet structure can be disposed on the back side of the knuckle 21 to form an antagonistic effect with the drive assembly. The sensing system includes a flexible sensor 6 integrated on or inside a non-Newtonian fluid core-shell structure 3, for real-time detection and feedback of detection and grasping status information.

[0023] In this embodiment, the dexterous hand includes five bionic fingers 2, such as Figure 1 As shown, the finger 2 on the left side of the figure is positioned on the side of the palm 1, simulating the position of the thumb of a human hand, and is connected to the palm 1 through the palm joint 231 at its base, realizing a single degree of freedom of vertical rotation. The other fingers 2 are connected to the palm 1 through their palm joints and the sliding bearing 221 of the palm 1, realizing a single degree of freedom of horizontal rotation, simulating the swinging function of the middle finger of a human hand. Specifically, each bionic finger 2 is composed of multiple joints 21 connected in series through a joint mechanism 22. Adjacent joints 21 are hinged through sliding bearings 221 and interphalangeal rivets 222, so that the finger 2 has the freedom of bending around the joint axis. The symmetrically arranged rotary bearings on the left and right sides of the finger 2 further ensure the flexibility and stability of the joint rotation.

[0024] Specifically, when the drive component tightens the tendon cord 4, the knuckle 21 bends around the joint mechanism 22, causing the non-Newtonian fluid core-shell structure 3 to contact the target object. At this time, the fluid is in a low-viscosity state, and the contact surface undergoes passive deformation to envelop the object's contour. When the sensing system detects a preset trigger condition, it adjusts the fluid state to change it to a high-viscosity state, rigidifying the contact surface to lock the grasping posture. After the drive component releases the tendon cord 4, the reset component uses elastic force to restore the knuckle 21 to its extended state. Compared with existing technologies, existing rigid contact surfaces cannot adapt to irregular objects, while this solution achieves controllable conversion between contact surface flexibility and rigidity through the non-Newtonian fluid core-shell structure 3. Existing elastomer contact surfaces rely on fixed elastic deformation, while this solution combines fluid state control and sensing feedback to dynamically adjust the mechanical properties of the contact surface according to the object's characteristics. This application can achieve flexible envelopment through low-viscosity fluid during the contact phase, reducing the risk of damage to vulnerable objects; and provides rigid support through fluid state switching during the grasping phase to prevent slippage.

[0025] By integrating a non-Newtonian fluid core-shell structure 3 on the fingertip of the bionic finger 2, the dexterous hand can automatically adjust the softness and stiffness of the contact surface according to the external force when contacting an object, achieving highly adaptive envelope grasping of objects of different shapes, sizes, and materials. When the external force is small or at the beginning of the grasp, the non-Newtonian fluid is in a low viscosity state, and the surface of the finger 2 is soft and deformable, which can effectively increase the contact area and disperse stress, thereby avoiding damage to fragile or flexible objects. When the grasping force increases or is disturbed by the external environment, the non-Newtonian fluid exhibits high viscosity or solid-like properties, and the contact surface becomes rigid and locked, which can enhance the grasping force and suppress slippage, achieving stable holding of the object. Thus, the dexterous hand can quickly switch between "flexible envelope" and "rigid support", significantly improving the grasping reliability under complex working conditions.

[0026] Furthermore, the present invention employs a design that combines tendon tractor 4 with elastic reset mechanism 5, enabling finger 2 to move flexibly and respond quickly. The structure is compact and has good mechanical controllability. The reset mechanism can automatically restore finger 2 to its initial posture after the driving force is released, improving operational efficiency and reducing system energy consumption. The non-Newtonian fluid core-shell structure 3 not only has high damping characteristics, effectively absorbing shock and vibration and preventing mechanical rebound, but also gives the entire finger 2 excellent vibration reduction and buffering capabilities, thereby improving the stability and safety of the grasping action.

[0027] In this embodiment, the flexible sealing shell of the non-Newtonian fluid core-shell structure 3 is a multi-layer composite structure, including at least an inner sealing layer, a middle reinforcing braided layer, and an outer wear-resistant layer to ensure its sealing performance, strength, and durability. The inner sealing layer is an elastic thin film layer that directly contacts the non-Newtonian fluid core 32, and can be made of thermoplastic polyurethane or silicone material. It is used to prevent fluid leakage and maintain the integrity of the sealed cavity. The middle reinforcing braided layer is a mesh structure layer woven from high-modulus fibers, and can be made of aramid fiber or carbon fiber braid. It is used to resist the tensile stress generated when the knuckle 21 is bent, preventing the shell from cracking. The outer wear-resistant layer is the outermost wear-resistant coating, and can be made of polyurethane composite material or rubber-based composite material. It is used to reduce frictional wear with the object surface during gripping, extending service life. The multi-layer composite structure design allows the core-shell contact surface to maintain flexibility while possessing higher structural strength and reliability, adapting to complex gripping environments and the requirements of multiple cycles of use. Among them, the outer surface of the non-Newtonian fluid core-shell structure 3 is covered with a flexible finger sleeve with anti-slip texture. The flexible finger sleeve with anti-slip texture refers to the concave and convex structure 33 layers covering the surface of the outer wear-resistant layer. Specifically, it can be achieved by molding or laser engraving process to form micron-level grooves on the surface of the silicone sleeve. This is used to increase the friction coefficient of the contact surface, which can improve the gripping force and stability when enveloping objects and prevent slippage or drop caused by surface smoothness or vibration. The finger sleeve also plays a secondary protective role, which can isolate the damage of sharp particles or liquids to the outer membrane of the core-shell.

[0028] As a preferred embodiment of the above, the non-Newtonian fluid core 32 is at least one of shear-thickening fluid (STF), magnetorheological fluid (MRF), or electrorheological fluid (ERF). Specifically, a shear-thickening fluid refers to a non-Newtonian fluid whose viscosity increases significantly with increasing shear rate, and can be implemented using a suspension system formed by dispersing silica particles in a polyethylene glycol-based liquid, which produces a hardening response when subjected to rapid extrusion or impact; a magnetorheological fluid refers to a carrier liquid dispersed with micron-sized magnetic particles, and can be implemented using a suspension formed by mixing carbonyl iron powder and silicone oil, whose rheological properties can be adjusted in real time by an external magnetic field; an electrorheological fluid refers to a carrier liquid dispersed with dielectric particles. The insulating oil can be a suspension formed by mixing barium titanate particles and mineral oil, and its viscosity change is controlled by an external electric field. Specifically, during the grasping process, when the non-Newtonian fluid core 32 uses a shear-thickening fluid, the squeezing action generated by the finger 2 contacting the object can induce local shear thickening of the fluid, thereby increasing the hardness of the contact area and enhancing the grasping stability. When magnetorheological fluid or electrorheological fluid is used, the overall viscosity of the core can be actively adjusted by applying an external magnetic field or electric field to achieve dynamic control of the grasping force. In the adaptive envelope stage, the core maintains a low viscosity state to allow flexible deformation of the fingertip, while in the stiffening and locking stage, the contact surface is stiffened by triggering a viscosity jump.

[0029] When the non-Newtonian fluid core 32 is a magnetorheological fluid (MRF) or an electrorheological fluid (ERF), a fluid state control system is also included. The fluid state control system includes a control device for applying a magnetic field or electric field to the non-Newtonian fluid core-shell structure 3. Specifically, the control device refers to the actuator that generates the magnetic field or electric field, which can be implemented using a miniature electromagnetic coil or electrode pair. The miniature electromagnetic coil or electrode pair is integrated into the bottom or periphery of the non-Newtonian fluid core-shell structure 3, and the fluid viscosity is precisely controlled by applying a local field strength. Specifically, when the non-Newtonian fluid core 32 is a magnetorheological fluid or an electrorheological fluid, the fluid state control system applies an external magnetic field or electric field to the core-shell structure through the control device, causing the particles or molecules inside the fluid to undergo directional arrangement or structural reorganization under the action of the physical field, thereby rapidly changing the apparent viscosity of the fluid.

[0030] In this embodiment, the control device is a miniature electromagnetic coil or electrode pair, integrated into the bottom or periphery of the non-Newtonian fluid core-shell structure 3. The miniature electromagnetic coil refers to a miniaturized coil structure capable of generating a local magnetic field, specifically achieved using multi-layered tightly wound copper wire combined with a ferromagnetic core. When energized, it can form a gradient magnetic field inside the non-Newtonian fluid core-shell structure 3, thereby altering the rheological properties of the magnetorheological fluid. Specifically, the electrode pair refers to a plate or mesh structure made of conductive material, specifically a composite of a flexible conductive polymer and a metal plating layer. When energized, it can form a uniform electric field inside the non-Newtonian fluid core-shell structure 3, thereby changing the viscosity distribution of the electrorheological fluid. The integration into the bottom or periphery refers to the integration of the control device... The core-shell structure 3 of the non-Newtonian fluid is embedded at the bottom edge or around its sidewall. Specifically, it can be seamlessly integrated with the flexible sealed shell using microelectronic packaging technology to maintain the integrity of the surface morphology of the knuckle 21. In specific applications where magnetorheological fluid or electrorheological fluid is used as the core, when it is necessary to trigger the fluid state switching, the micro electromagnetic coil generates a controllable magnetic field through pulse current excitation, causing the particles inside the magnetorheological fluid to form a chain structure to increase the viscosity. The electrode pair forms an electric field inside the fluid by applying voltage, forcing the particles in the electrorheological fluid to polarize and form a columnar arrangement to enhance rigidity. The spatial layout of the control device is optimized. Bottom integration can avoid affecting the tactile sensation of the fingertip, while peripheral integration can improve the coverage uniformity of the magnetic field or electric field.

[0031] This invention also includes a control method for the biomimetic flexible adaptive dexterous hand as described above, such as... Figure 7 As shown, the method includes: A10: Adaptive Envelope Stage: Control the dexterous hand to make the gripping surface with the non-Newtonian fluid core-shell structure 3 contact the object at a predetermined contact speed, and maintain the non-Newtonian fluid core-shell structure 3 in a low viscosity state, so that the gripping surface produces flexible deformation to passively fit and envelop the outline of the object. A20: Rigidification and Locking Stage: When the preset triggering conditions are met, the non-Newtonian fluid core-shell structure 3 is controlled to switch from a low viscosity state to a high viscosity state, thereby increasing its apparent viscosity and making the gripping surface rigid to lock the gripping posture.

[0032] To achieve adaptive deformation and rigidification locking during the grasping process of objects of different shapes and materials by a dexterous hand, a phased control strategy is proposed, mainly consisting of two core steps: an adaptive envelope phase and a rigidification locking phase. In the adaptive envelope phase, the control system first receives grasping target information from the upper-level task planning module. Based on the object's geometric contour, material properties, and contact sensitivity level, it sets initial parameters for contact speed, finger-to-finger angle, and tension of the tendon 4. Subsequently, the drive component controls the tendon 4 to tighten at a low rate, causing the surface of the non-Newtonian fluid core-shell structure 3 at the finger 2 to gradually contact the object surface in a flexible state. During this phase, the non-Newtonian fluid maintains a low viscosity state, and its flexible sealing shell can undergo local deformation under external force, thereby achieving the binding of the object surface. Dynamic contact and contour envelopment enable the dexterous hand to stably envelop the target object without applying excessive contact force, avoiding slippage or damage caused by hard contact. During the rigidification and locking phase, when the system detects that the preset trigger conditions are met, the control module activates the non-Newtonian fluid state regulation mechanism, which switches the fluid from a low-viscosity flexible state to a high-viscosity rigid state. This switch can be triggered by physical excitation (such as applying an external field) or mechanical disturbance (such as driving the tendon rope 4 to tighten slightly and quickly), thereby instantly increasing the apparent viscosity of the non-Newtonian fluid and causing a chain aggregation effect in the internal fluid molecular structure, which significantly enhances the stiffness of the core-shell structure. At this time, the gripping surface of the dexterous hand changes from a flexible form to a rigid state, which can effectively fix the current gripping posture and prevent the object from slipping during subsequent handling or disturbance.

[0033] Compared with existing technologies, traditional control methods rely solely on the deformation recovery characteristics of elastic materials and cannot actively adjust the stiffness of the gripping surface. This can lead to damage when gripping fragile objects due to excessive contact force, or slippage when handling smooth objects due to insufficient damping. This method uses a non-Newtonian fluid state switching mechanism to maintain low stiffness during the envelope phase to achieve flexible contact, and increases stiffness during the locking phase to enhance gripping stability. This solves the problem that a single material property cannot simultaneously achieve both flexibility and load-bearing capacity.

[0034] In this embodiment, the non-Newtonian fluid core-shell structure 3 is switched to a high viscosity state in any of the following ways: When the non-Newtonian fluid core 32 is a shear-thickening fluid, a high viscosity state is triggered by the slight tightening action of the drive component; When the non-Newtonian fluid core 32 is a magnetorheological fluid or an electrorheological fluid, high viscosity switching is achieved by applying an external magnetic field or electric field through the fluid state control system.

[0035] Specifically, a method is proposed to control the switching of the non-Newtonian fluid core-shell structure 3 to a high viscosity state during the stiffening and locking stage. This includes two implementation approaches: For shear-thickening fluids, the driving component is controlled to perform a rapid micro-tightening action to increase the fluid shear rate and trigger the shear-thickening effect; For magnetorheological or electrorheological fluids, the fluid state control system is controlled to apply an external magnetic field or electric field. The rapid micro-tightening action refers to the driving component performing a millimeter-level tightening operation on the tendon 4 within milliseconds. By precisely controlling the tightening amplitude and speed, the internal shear rate of the fluid is increased while maintaining a stable gripping posture. Applying an external magnetic field or electric field refers to generating a magnetic field strength of 0.1-1T through a micro electromagnetic coil or applying an electric field strength of 100-1000V / mm through electrodes. This can be achieved using a patch electromagnetic array or an embedded electrode structure, thereby changing the microscopic particle arrangement state of the fluid to achieve a viscosity jump.

[0036] In this embodiment, when grasping fragile objects, when the flexible sensor 6 detects that the pressure on the fingertip reaches a set threshold, if the core is a shear-thickening fluid, the drive motor immediately performs a 0.5mm stroke, 50ms duration, tightening action of the tendon 4, causing the knuckle 21 to vibrate slightly, thereby stimulating the fluid shear-thickening effect; if the core is a magnetorheological fluid, the micro electromagnetic coils arranged around the knuckle 21 are activated, establishing a directional magnetic field within 5ms to form a chain-like structure of ferromagnetic particles. Both methods can maintain the shape of the grasping contact surface while increasing the fluid viscosity by two orders of magnitude, achieving a seamless switch from flexible envelope to rigid locking.

[0037] As a preferred embodiment of the above, such as Figure 8 As shown, the triggering conditions are determined through the following steps: A21: Real-time acquisition of multimodal sensing data provided by the sensing system, including pressure distribution, strain distribution and vibration signals; A22: Extracting temporal features from multimodal sensing data; A23: Input the temporal features into the pre-trained stability evaluation model to obtain the slip probability of the current grasping state; A24: When the probability of slippage exceeds the dynamic threshold, the trigger condition is determined to be met.

[0038] First, a multi-modal sensing system deployed on the surface of the dexterous hand collects multi-source data in real time, including pressure distribution, strain distribution, and vibration signals. After sampling, filtering, and normalization, a multi-dimensional time-series input matrix is ​​formed. Then, a sliding time window mechanism is used to extract time-series features such as pressure center offset, strain gradient change rate, root mean square value of vibration acceleration and its derivative, cross-correlation coefficient, and power spectral density from the data. These features are then input into a pre-trained stability evaluation model. This model preferably adopts a deep neural network structure based on Long Short-Term Memory (LSTM) and combines an attention mechanism to enhance the temporal weight expression of key features, thereby accurately calculating the slip probability in the current grasping state. When the calculated slip probability exceeds the dynamic threshold, the system automatically determines that the current grasping is in a potentially unstable state and sends a control command to the drive component, triggering the non-Newtonian fluid core-shell structure 3 to switch from a low-viscosity state to a high-viscosity state, realizing the stiffening and locking of the grasping surface. The dynamic threshold is adaptively adjusted according to historical states and real-time feedback signals to ensure the sensitivity and stability of the response.

[0039] As a preferred embodiment of the above, in step A23, the temporal features are input into a pre-trained stability evaluation model to obtain the slip probability of the current grasping state. The steps include: A231: Multimodal sensing data is segmented according to a preset time window, and pressure, strain, and vibration signals are synchronized in time; A232: Extract contact area change rate, normal force fluctuation, strain gradient and spectral energy characteristics from time synchronization data, and normalize the characteristics; A233: Input the processed features into the temporal coding module to generate temporal embedding vectors; A234: Input the temporal embedding vector into the pre-trained stability evaluation model and output the slip probability of the current grasping state; A235: Smooths the slip probability and compares it with a dynamic threshold that is adaptively updated based on the normal force and friction coefficient. When the threshold is exceeded continuously, a trigger signal is output.

[0040] First, a multimodal sensor array is used to continuously acquire pressure signals from the grasping contact surface, strain signals from the fingertips and palm, and micro-vibration signals. The acquired multi-source data is time-aligned using a unified clock synchronization mechanism and stored in segments according to preset time windows for subsequent feature analysis and dynamic modeling. Within each time period, the control module calculates indicators such as the rate of change of contact area, normal force fluctuation amplitude, strain distribution gradient, and frequency domain energy characteristics to characterize the contact state and stability features between the object and the flexible surface during the grasping process. To further improve the temporal integrity and nonlinear mapping capability of feature representation, this invention designs a temporal encoding module to extract deep temporal dependencies from multidimensional feature sequences. Preferably, this module can adopt a Long Short-Term Memory (LSTM) network structure, using input gates, forget gates, and output gates to process the temporal information. Alternatively, a Transformer encoder structure can be adopted to model the temporal correlation between different features using a multi-head self-attention mechanism. After encoding, the output embedding vector can represent the dynamic evolution trend of the grasping state in a high-dimensional feature space. Subsequently, the temporal embedding vector is input into a stability evaluation model pre-trained on large-scale experimental data. This model obtains the ability to nonlinearly map from input features to slip probability through supervised learning, and can output the slip probability value representing the risk level of the current grasping state in real time. To avoid misjudgment caused by instantaneous noise or short-term fluctuations, the slip probability sequence is time-smoothed in the output stage. A moving average or exponentially weighted moving average algorithm is used to reduce high-frequency jitter. At the same time, the dynamic threshold is adaptively updated according to the current normal force level, the friction characteristics of the object surface, and historical state data to reflect the real-time grasping conditions. When the smoothed slip probability value continuously exceeds the threshold, the controller triggers a stiffening control signal, driving the non-Newtonian fluid core-shell structure 3 into a high-viscosity state, thereby enhancing grip stability. In a specific embodiment, when the dexterous hand is carrying a cylindrical glass bottle, the sensor array detects changes in normal force fluctuations and strain gradients. The slip probability output by the evaluation model increases and exceeds the dynamic threshold. After smoothing confirmation, the system automatically executes stiffening control, causing the surface of the finger 2 to quickly become rigid, preventing the bottle from slipping. Through this scheme, the present invention can realize an intelligent judgment mechanism based on multimodal temporal features and deep learning models, enabling the dexterous hand to have adaptive grasping stability control capabilities, significantly improving operational accuracy and anti-interference performance.

[0041] In this embodiment, after completing the rigidification and locking and entering the object transport process, the method further includes: A30: Dynamic stabilization phase: During the transport of the object, the dynamic signal of vibration acceleration or pressure distribution fluctuation is monitored in real time by the sensing system; Based on dynamic signals, the viscosity of the non-Newtonian fluid core-shell structure 3 is dynamically adjusted in real time to actively suppress slippage or vibration.

[0042] The dynamic stabilization phase refers to the continuous dynamic feedback adjustment process after the object's grasping posture is locked. This can be achieved by forming a closed-loop control circuit with the embedded controller and the sensing system to cope with dynamic instability caused by external disturbances during transportation. The dynamic signal refers to the vibration acceleration or contact pressure distribution change data collected by the flexible sensor 6. This can be measured using a piezoresistive sensor or a capacitive sensor, and its data characteristics can reflect the relative motion trend between the object and the grasping surface. Real-time dynamic adjustment refers to the continuous or intermittent adjustment of the viscosity of the non-Newtonian fluid core-shell structure 3 based on the sensor feedback signal. This can be achieved by changing the magnetic field strength, electric field strength, or mechanical vibration frequency to match the rheological properties of the fluid core with the current transportation conditions. The pulsating application of the control field refers to the application of a periodically changing field strength to the non-Newtonian fluid core-shell structure 3. This can be achieved by using a waveform generator to control the power supply current of the electromagnetic coil or electrodes, and by frequency matching and phase adjustment to actively cancel the vibration energy.

[0043] Specifically, after the rigidification and locking are completed, the flexible sensor 6 continuously collects the vibration acceleration signal of the contact area between the gripping surface and the object. When a vibration component exceeding a set threshold is detected, the control unit analyzes the signal spectrum and identifies the dominant frequency component. Subsequently, the control device applies a magnetic field or electric field with the same pulsating frequency as the dominant frequency, causing the non-Newtonian fluid core-shell structure 3 to produce periodic viscosity changes, forming a damping force opposite to the vibration direction. For example, when the object tends to slide laterally due to external impact, the pressure distribution sensor detects a decrease in the pressure value at the edge of the contact area. At this time, the local viscosity is increased by enhancing the magnetic field strength in the corresponding area, thereby increasing the frictional resistance between the gripping surface and the object.

[0044] The real-time dynamic adjustment is achieved by applying a pulsating control field, and the frequency of the pulsation is related to the dominant frequency of the dynamic signal, and is controlled in an anti-phase manner to counteract the vibration.

[0045] Specifically, by applying periodically varying magnetic, electric, or mechanical shear forces to the non-Newtonian fluid core-shell structure 3, its viscosity pulsates over time, and the pulsation frequency is correlated with the dominant frequency of the vibration signal detected by the sensing system. This anti-phase method cancels out vibration energy, thereby actively suppressing slippage or vibration caused by external disturbances. Specifically, the sensing system uses an accelerometer to collect vibration signals in real time, and the embedded controller extracts the dominant frequency component using a Fast Fourier Transform (FFT) and generates a control signal corresponding to this dominant frequency but with an opposite phase based on a Phase-Locked Loop (PLL) algorithm. For the non-Newtonian fluid core-shell structure 3 using magnetorheological fluid, the controller adjusts the coil current using PWM to generate a magnetic field pulsation signal with the same vibration dominant frequency but a 180° phase reversal, causing the yield stress of the magnetorheological fluid to change periodically over time, thus suppressing slippage or vibration caused by external disturbances. The system achieves stiffening during vibration and release during reverse vibration to counteract external disturbances. For electrorheological fluid structures, a similar anti-phase control is achieved by adjusting the pulsating electric field signal output by the high-voltage drive module, with a preferred frequency range of 10–200 Hz and a modulation depth of ±30%–50%. For shear-thickening fluid structures, a micro servo driver controls the tendon rope 4 to perform micro-stretching actions in a synchronous but anti-phase manner, inducing periodic changes in the shear rate within the fluid to form controllable damping. The system adopts a multi-channel closed-loop control architecture, calculates the vibration frequency, amplitude, and phase in real time, and dynamically adjusts the amplitude and phase matching of the control field to achieve rapid adaptive adjustment with a millisecond-level response period. By achieving a rheological response opposite to the external vibration in the time domain, active damping compensation is effectively formed, enabling the dexterous hand to maintain high precision and high stability gripping performance in complex dynamic environments.

[0046] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A biomimetic flexible adaptive dexterous hand, characterized in that, include: The hand, multiple bionic fingers, drive components, reset components, and sensing system; The bionic finger includes multiple phalanges connected in series by a joint mechanism, and the bionic finger is movably connected to the palm through the phalangeal joint; At least a portion of the finger joints have a non-Newtonian fluid core-shell structure on the fingertip side, the non-Newtonian fluid core-shell structure comprising a closed cavity formed by a flexible sealing shell and a non-Newtonian fluid core filling the closed cavity. The drive assembly includes a tendon cord and a drive component that pass through the internal channel of the knuckle, and drives the bending motion of the bionic finger by extending and retracting the tendon cord. The reset assembly includes an elastic reset mechanism disposed on the back side of the knuckle, which is used to restore the knuckle to an extended state after the driving force of the driving assembly is released. The sensing system includes a flexible sensor integrated on or inside the non-Newtonian fluid core-shell structure for real-time detection and feedback of the detection and grasping status information.

2. The biomimetic flexible adaptive dexterous hand according to claim 1, characterized in that, The flexible sealing shell of the non-Newtonian fluid core-shell structure is a multi-layer composite structure, including at least an inner sealing layer, an intermediate reinforcing braided layer, and an outer wear-resistant layer.

3. The biomimetic flexible adaptive dexterous hand according to claim 1, characterized in that, The non-Newtonian fluid core is at least one of shear-thickening fluid, magnetorheological fluid, or electrorheological fluid.

4. The biomimetic flexible adaptive dexterous hand according to claim 3, characterized in that, When the non-Newtonian fluid core is a magnetorheological fluid or an electrorheological fluid, it also includes a fluid state control system, which includes a control device for applying a magnetic field or an electric field to the non-Newtonian fluid core-shell structure.

5. A control method for a biomimetic flexible adaptive dexterous hand as described in claim 1, characterized in that, The method includes: Adaptive envelope stage: Control the dexterous hand to make the gripping surface with the non-Newtonian fluid core-shell structure contact the object at a predetermined contact speed, and maintain the non-Newtonian fluid core-shell structure in a low viscosity state, so that the gripping surface produces flexible deformation to passively fit and enclose the object contour. Rigidification and locking phase: When the preset triggering conditions are met, the non-Newtonian fluid core-shell structure is controlled to switch from the low viscosity state to the high viscosity state, thereby increasing its apparent viscosity and making the gripping surface rigid to lock the gripping posture.

6. The control method for the biomimetic flexible adaptive dexterous hand according to claim 5, characterized in that, The non-Newtonian fluid core-shell structure is controlled to switch to a high viscosity state in any of the following ways: When the non-Newtonian fluid core is a shear-thickening fluid, a high viscosity state is triggered by the slight tightening action of the drive component; When the non-Newtonian fluid core is a magnetorheological fluid or an electrorheological fluid, high viscosity switching is achieved by applying an external magnetic field or electric field through a fluid state control system.

7. The control method for the biomimetic flexible adaptive dexterous hand according to claim 5, characterized in that, The triggering condition is determined through the following steps: Real-time acquisition of multimodal sensing data provided by the sensing system, including pressure distribution, strain distribution and vibration signals; Extract temporal features from the multimodal sensing data; The temporal features are input into a pre-trained stability evaluation model to obtain the slip probability of the current grasping state; When the slippage probability exceeds the dynamic threshold, the triggering condition is determined to be met.

8. The control method for the biomimetic flexible adaptive dexterous hand according to claim 7, characterized in that, The temporal features are input into a pre-trained stability evaluation model to obtain the slip probability of the current grasping state. The steps include: The multimodal sensing data is segmented according to a preset time window, and the pressure, strain, and vibration signals are synchronized in time. The contact area change rate, normal force fluctuation, strain gradient and spectral energy characteristics are extracted from the time synchronization data, and the characteristics are normalized. The processed features are input into the temporal coding module to generate temporal embedding vectors; The temporal embedding vector is input into a pre-trained stability evaluation model, which outputs the slip probability of the current grasping state. The slip probability is smoothed and compared with a dynamic threshold that is adaptively updated based on the normal force and friction coefficient. When the slip probability exceeds the threshold continuously, a trigger signal is output.

9. The control method for the biomimetic flexible adaptive dexterous hand according to claim 5, characterized in that, After completing rigidification and locking and entering the object transport process, it also includes: Dynamic stabilization phase: During the transport of the object, the dynamic signals of vibration acceleration or pressure distribution fluctuations are monitored in real time through the sensing system. Based on the dynamic signal, the viscosity of the non-Newtonian fluid core-shell structure is dynamically adjusted in real time to actively suppress slippage or vibration.

10. The control method for the biomimetic flexible adaptive dexterous hand according to claim 9, characterized in that, The real-time dynamic adjustment is achieved by applying a pulsating control field, and the frequency of the pulsation is related to the dominant frequency of the dynamic signal and controlled in an anti-phase manner to counteract the vibration.

Images