A bionic self-sensing tendon for a wire-driven dexterous hand, a bionic dexterous hand, and a control method thereof.

Abstract

This invention relates to a bionic self-sensing tendon for a wire-driven dexterous hand, a bionic dexterous hand, and a control method. It employs an integrated molding process of MWCNT / PDMS nano-conductive composite material and a flexible elastic matrix, enabling the tendon to perform traditional tendon-wire transmission functions while possessing real-time self-sensing capabilities for tensile loads. By placing the bionic self-sensing tendon within an underactuated wire-driven dexterous hand, and combining it with an end-effector flexible capacitive pressure-sensitive sensing unit, a dual-redundant complementary sensing system for tension and pressure is constructed. This system can automatically switch sensing modes in scenarios such as end-effector contact failure, posture deviation, and grasping irregular targets, significantly improving the force control stability and operational reliability of tendon-driven equipment such as bionic dexterous hands. It solves the industry problems of unstable fingertip contact and easy sensing failure in underactuated dexterous hands, providing an integrated and universal solution for high-precision compliant force control, safe drive protection, and stable interactive operation.

Description

Technical Field

[0001] This invention belongs to the field of biomimetic sensing, flexible actuation and robot perception technology, and relates to a biomimetic self-sensing tendon, a biomimetic dexterous hand and control method for a line-driven dexterous hand. Specifically, it relates to a self-sensing tendon, a biomimetic dexterous hand and control method based on the human Golgi tendon organ, and is applicable to tendon-line driven biomimetic dexterous hands, rehabilitation devices, humanoid robot joints and flexible actuators. Background Technology

[0002] In tendon-driven bionic mechanisms (such as dexterous hands, exoskeletons, and rehabilitation robots), the key to achieving precise force control and safe interaction lies in the effective sensing of tendon tension. However, existing mainstream force sensing solutions have significant limitations: on the one hand, contact sensing relying on single-point tactile sensation of the fingertips or arrayed electronic skin has inherent defects. The former is prone to sensing failure due to contact offset or angular deviation and cannot reflect the actual load at the driving end; the latter, although covering a wide area, faces problems such as complex manufacturing processes, cumbersome wiring, high costs, and difficult integration, which seriously restricts the lightweighting and mass application of the device. In addition, traditional tendons only serve as passive transmission components and lack proprioceptive sensing capabilities, making it impossible for the system to monitor the tendon's own tension, relaxation, overload, and fatigue state, resulting in the transmission chain operating in a "black box" state for a long time.

[0003] More importantly, existing technologies generally suffer from the drawbacks of limited sensing modalities and poor robustness. Whether using a single end-effector contact sensor or an external tension / compression sensor, in irregular grasping or dynamic motion in unstructured environments, signal loss can easily lead to control oscillations or even grasping failure. This "end-effector-heavy, actuator-light" sensing architecture fails to establish a collaborative feedback mechanism between the actuator and the contact end. It cannot cope with signal fluctuations under complex operating conditions, nor can it trigger immediate protection in the event of overload, fundamentally differing from the highly robust force sensing mechanism naturally possessed by the human body.

[0004] In contrast, biological mechanisms reveal that the Golgi tendon organ at the junction of tendons and muscles in the human body can directly sense muscle tension and trigger protective reflexes, serving as a model of efficient force control and safety protection. Unfortunately, current technologies have not yet overcome the dual bottlenecks of materials and architecture: neither a low-cost, integrated, and directly replaceable self-sensing bionic tendon has been developed, nor is a universal sensing architecture that mimics biological neural pathways and integrates signals from the driving and contact ends. Therefore, an innovative solution that dually replicates biological mechanisms at both the structural and sensory levels is urgently needed to fill the gaps in existing technologies and propel bionic mechanisms from the laboratory to practical engineering applications. To this end, this invention proposes a bionic self-sensing tendon for a wire-driven dexterous hand, using nanocomposite materials to achieve self-sensing within the tendon body, complemented by end-effector pressure-sensitive sensing for redundant integration, achieving both structural and sensory replication and engineering implementation. Summary of the Invention

[0005] In view of this, in order to solve the shortcomings of existing force sensing schemes of tendon-driven bionic mechanisms, such as strong contact dependence, high cost and difficult integration, lack of proprioception at the driving end, and poor single-modal robustness, the present invention provides a bionic self-sensing tendon, a bionic dexterous hand, and a control method for a line-driven dexterous hand, in order to solve the problems of difficult support for precise force control and safe interaction in complex scenarios. The present invention has failed to achieve a low-cost integrated self-sensing bionic tendon and lacks a universal sensing architecture for collaboration between the driving end and the contact end.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A biomimetic self-sensing tendon for a wire-driven dexterous hand includes an elastic tube made of a flexible elastic matrix, a nano-conductive composite material core layer filled inside the elastic tube, and port fixing molds fixed to both ends of the elastic tube. The outer end of the port fixing mold has a through channel for a polymer polyester tendon wire and a copper wire at its center. After the polymer polyester tendon wire and the copper wire are introduced, they are fixed to one end by spot welding. The copper wire passes through the port fixing mold and connects to the nano-conductive composite material core layer to form an electrode. The tendon wire is connected to the port fixing mold to transmit tensile force.

[0008] A bionic dexterous hand based on bionic self-sensing tendons includes four fingers, each with independently configured modules capable of flexion and extension movements, and a thumb capable of flexion, extension, and lateral swing movements. The four fingers are driven independently by four linear servos, the thumb is controlled by a lateral swing servo for lateral swing, and the linear servos control flexion and extension. The wrist is controlled by a motor to rotate a synchronous belt, causing the dexterous hand to bend. The flexion and extension movements of the four fingers and thumb are all achieved through tendon-wire transmission. Pressure-sensitive sensors capable of sensing pressure are installed at the fingertips of the four fingers and thumb. This bionic dexterous hand is equipped with the aforementioned bionic self-sensing tendons, which replace a section of tendon on the back of the hand to detect the driving force of the tendon. Combined with the flexible pressure-sensitive sensors at the fingertips, a dual-sensory feedback is formed, enabling precise control of gripping force.

[0009] The control method for a bionic dexterous hand based on bionic self-sensing tendons is characterized by the following steps:

[0010] S1. The host computer detects the precise coordinates, shape and size information of the object under inspection through the vision module. Based on the object under inspection, the host computer gives appropriate pre-grabbing gestures and pre-grabbing forces and feeds them back to the dexterous hand control system through data packets.

[0011] S2. The host computer sends motion commands to the robotic arm via TCP communication, plans the motion path based on the pose information of the object being inspected, and controls the robotic arm to move the dexterous hand to the predetermined grasping pose of the object being inspected.

[0012] S3, the dexterous hand control system parses the received data packet, obtains the pre-grasping gesture and pre-grasping force, drives the thumb lateral swing joint and the four finger flexion and extension joints to adjust to the initial grasping posture, and the bionic tendon tension sensor and fingertip pressure sensor enter the real-time sensing ready state.

[0013] S4. The dexterous hand begins to grasp according to the pre-grasping gesture, while the fingertip pressure sensor and the tension sensor work together to sense and control to ensure stable grasping of the object being inspected.

[0014] The beneficial effects of this invention are as follows:

[0015] 1. The bionic self-sensing tendon for a wire-driven dexterous hand disclosed in this invention breaks through the limitations of the traditional tendon-driven system's "transmission-sensing separation" architecture, natively integrating sensing functions into the tendon body: using a rubber hose as a flexible substrate and MWCNT / PDMS nano-conductive composite material as a functional core layer, the bionic self-sensing tendon, in addition to undertaking the core transmission functions of transmitting tension and driving finger flexion and extension, can simultaneously convert tensile deformation within the 0-20N range into resistance change signals, directly outputting real-time tension data. This design completely eliminates the need for external tension and compression sensors, additional mounting brackets, and complex wiring, avoiding the occupation of internal space in the dexterous hand by additional structures, fundamentally simplifying the system topology, and significantly improving the integration and miniaturization of the tendon-driven mechanism.

[0016] 2. The bionic dexterous hand based on bionic self-sensing tendons disclosed in this invention addresses the industry pain points of underactuated dexterous hands, such as fingertips easily dangling and single-point tactile sensor failure when grasping irregular objects. This invention constructs a dual-redundant sensing system of "drive-end tension - contact-end pressure": when grasping regular objects, the dual signals mutually verify each other to improve force control accuracy; when the fingertips are not in contact with the object or the pressure-sensitive data is invalid, the system can automatically switch to tendon tension control mode, indirectly calculating the grasping force based on the static mapping relationship, completely avoiding the risk of disconnection in a single sensing mode. Experimental verification shows that this collaborative mechanism increases the success rate of grasping typical objects such as cola cans and bananas from 75% and 55% to 100% and 90%, respectively, and can maintain sensing continuity and grasping stability even under complex working conditions such as dynamic adjustment of grasping posture and object sliding.

[0017] 3. The bionic dexterous hand based on bionic self-sensing tendons disclosed in this invention completely replicates the physiological protection mechanism of the human Golgi tendon organ: the bionic tendon can monitor the load of the transmission chain in real time. Once the tensile force exceeds the safety threshold of 20N, the system will trigger overload protection in milliseconds, automatically reducing the output of the drive unit or directly cutting off the power to avoid tendon rupture and joint structure damage. At the same time, based on the dual-end sensing compliant force control strategy, the gripping force can be adaptively adjusted according to the stiffness of the object—automatically reducing the expected force for flexible objects such as oranges and bananas to avoid squeezing damage, and matching sufficient clamping force for rigid workpieces to prevent slippage. This not only ensures the safety of operation, but also reduces the fatigue wear of transmission components and significantly extends the service life of the entire system.

[0018] 4. The bionic dexterous hand based on bionic self-sensing tendons disclosed in this invention uses industrially available and commonly used consumables: multi-walled carbon nanotubes, PDMS matrix, and rubber tubing are readily available in the market and have extremely low costs, eliminating the need for expensive semiconductor processing equipment or specially customized materials. The preparation process only requires conventional processes such as ultrasonic dispersion, mechanical stirring, vacuum degassing, and room temperature curing, without the need for complex procedures such as photolithography, micro-nano processing, and array wiring. The preparation cycle for a single sensor is no more than 3 hours, with a yield rate consistently above 90%. The cost per unit is only 1 / 10 of that of commercially available tensile sensors with similar performance, making it very suitable for small- to medium-batch production and significantly reducing the threshold for sensor modification in bionic dexterous hands.

[0019] 5. The bionic self-sensing tendon for wire-driven dexterous hands disclosed in this invention is highly compatible with traditional polymer tendon threads in terms of size and mechanical properties. It can be directly embedded into various tendon-driven devices without modifying the original transmission structure. It can be used for precise force control in multi-degree-of-freedom bionic dexterous hands, integrated into rehabilitation exoskeletons and assistive robotic arms to monitor human-machine interaction forces, and deployed in flexible joints and bionic robot leg transmission chains to achieve overload warnings. Its lightweight and low-power characteristics are suitable for embedded system integration requirements. It does not require additional dedicated data acquisition equipment and can be quickly ported to tendon-driven systems in different scenarios. It is widely compatible with humanoid robot transmission systems such as bionic dexterous hands, rehabilitation exoskeletons, flexible driven joints, and medical rehabilitation devices, and has extremely strong cross-domain adaptability.

[0020] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0022] Figure 1 This is a schematic diagram of the back structure of the self-sensing tendon-driven dexterous hand of the present invention;

[0023] Figure 2 This is a schematic diagram of the front structure of the self-sensing tendon-driven dexterous hand of the present invention.

[0024] Figure 3 This is a schematic diagram of the structure of the index finger and thumb of the dexterous hand in this invention;

[0025] Figure 4 This is a schematic diagram of the biomimetic self-sensing tendon in this invention;

[0026] Figure 5 This is a flowchart illustrating the preparation process of the nano-conductive composite material core layer in this invention;

[0027] Figure 6 This is a simplified diagram of the coordinates of the index finger joint in the dexterous hand in this invention;

[0028] Figure 7 This is a simplified diagram of the thumb joint coordinates in the dexterous hand of this invention;

[0029] Figure 8 This is a flowchart of the tensile and compressive force sensing process in this invention.

[0030] Figure 9 This is a schematic diagram of the movement of the index finger in this invention;

[0031] Figure 10 This is a schematic diagram of the force balance of the index finger in this invention;

[0032] Figure 11 This is a diagram of the dexterous hand pinching with two fingers in this invention;

[0033] Figure 12 This is a diagram of the dexterous hand grasping with three fingers in this invention;

[0034] Figure 13 This is a diagram showing the dexterous finger-side pinching technique used in this invention;

[0035] Figure 14 This is a diagram of the five-finger envelope grasping technique in this invention;

[0036] Figure 15 This is a flowchart of the dexterous hand system for grasping oranges in this invention;

[0037] Figure 16 This is a flowchart of the dexterous hand system in this invention for grasping a cola can;

[0038] Figure 17 This is a flowchart of the dexterous hand system for grasping bananas in this invention.

[0039] Reference numerals: 1-Port fixing mold; 2-Bionic self-sensing tendon; 3-Tendon line; 4-Synchronous belt; 5-Linear servo; 6-Side swing servo; 7-Reset torsion spring; 8-Copper wire; 9-Nano-conductive composite core layer; 10-Rubber hose; 11-Distal phalanx; 12-Middle phalanx; 13-Proximal phalanx; 14-Thumb distal phalanx; 15-Thumb proximal phalanx; 16-Side swing joint. Detailed Implementation

[0040] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention.

[0041] The structure of the underactuated, line-driven bionic dexterous hand with bionic self-sensing tendons is as follows:

[0042] like Figure 1 , 2 The example shown here employs a lightweight, underactuated, wire-driven bionic dexterous hand. The overall design utilizes a tendon-wire transmission system, which not only simulates the function of human hand tendons but also offers advantages over linkage-based dexterous hands, such as better self-adaptability and lighter weight. Furthermore, it provides greater gripping force compared to flexible grippers. It features six active degrees of freedom, balancing a bionic shape with everyday grasping capabilities. The dexterous hand adopts a modular underactuated structure, using tendon-wire transmission as the drive method and a linear servo motor 5 as the actuator. Taking the index finger as an example, as shown in the attached diagram... Figure 3 As shown, it includes three phalanges: the distal phalanx 11, the middle phalanx 12, and the proximal phalanx 13. A return torsion spring 7 is embedded at all rotational joints. Its purpose is to allow the fingers to autonomously return to an extended state when not bent. By simulating the grasping mechanism of human hand muscles and tendons, this paper places the actuator within the lower forearm structure of the dexterous hand to mimic the lateral forearm muscles. One end of the tendon is fixed to the actuator end, and the other end is fixed to the fingertip. The structural characteristics of each phalanx limit the joint rotation angle, making the joint angle 0-90° towards the palm. Combined with the aforementioned tendon actuator and torsion spring design, it achieves flexion and extension movements of all four fingers. (See attached diagram) Figure 3 As shown in the diagram of the thumb structure, unlike the four-finger structure, the thumb has only two phalanges: the distal phalanx 14 and the proximal phalanx 15. However, it still has three degrees of freedom, including flexion and extension of the distal and proximal phalanges, flexion and extension of the proximal phalanx and metacarpal joints, and the lateral swing joint's outward and inward movement of the entire finger towards the palm, with the palm as the axis of rotation. The lateral swing joint is designed to be directly driven by a servo motor, enabling rotation of 0-110° towards the palm. The other two joints, similar to those of the four fingers, are driven by a single actuator that pulls a tendon to achieve flexion and extension of the thumb.

[0043] The specific structure and fabrication process of the biomimetic self-sensing tendon are as follows:

[0044] As attached Figure 4 As shown, the biomimetic self-sensing tendon 2 includes: a flexible elastic matrix, a nano-conductive composite material core layer 9, and a port fixing mold 1 for end fixation. The nano-conductive composite material core layer 9 is located in an elastic hose made of the flexible elastic matrix, and the port fixing mold 1 is located at both ends of the elastic hose made of the flexible elastic matrix.

[0045] The flexible elastic matrix is ​​made of rubber hose 10 with a modulus between 0.5 and 2.0 GPa, which is similar to the modulus of human tendon. The inner wall of rubber hose 10 has good hydrophilicity, which makes it easy to bond with composite materials. Using rubber hose 10 as the flexible tensile matrix of sensor can realize effective stretching and resistance change under tension, while protecting the material itself and providing stable tensile deformation and elastic rebound capability.

[0046] As attached Figure 5 As shown, the preparation method of the nano-conductive composite core layer 9 is as follows: Multi-walled carbon nanotubes (MWCNTs) are weighed as conductive fillers; component A (main body) and component B (curing agent) of polydimethylsiloxane (PDMS) are prepared, along with n-hexane as a dispersant; simultaneously, rubber tubing 10 with an inner diameter of 1.5 mm and an outer diameter of 3.5 mm is cut as encapsulation material, and polymerized polyester thread (as the tendon body), wires, syringes, an ultrasonic oscillator, a mechanical stirrer, and a vacuum drying oven are prepared; the weighed MWCNTs (preferably 7% by mass) are added to 30 mL of n-hexane and oscillated in an ultrasonic oscillator for 20 minutes to fully disperse them into a suspension; subsequently, 10 g of PDMS-A component is added to the suspension and stirred with a mechanical stirrer at 200 °C. Stir continuously at a speed of r / min for 120 minutes to uniformly disperse MWCNT in the PDMS matrix and allow n-hexane to completely evaporate; add PDMS-B curing agent to the above mixture at a mass ratio of PDMS-A to PDMS-B of 40:3 (i.e., 10g:0.75g), and continue stirring at a speed of 200 r / min for 10 minutes to ensure uniform mixing; then, place the mixed slurry in a vacuum drying oven and perform 3 complete vacuum cycles to thoroughly remove air bubbles introduced during the stirring process, obtaining a bubble-free, uniform conductive slurry.

[0047] In this invention, the method for fabricating the end fixation and electrode lead-out structure is as follows: A syringe is used to draw up the degassed MWCNT / PDMS conductive slurry and slowly and continuously inject it into the prepared rubber tubing 10 until it is full. Before the slurry solidifies, a polymer polyester tendon wire 3 is quickly inserted from one end of the rubber tubing 10 as the tendon core and exits from the other end. Simultaneously, two copper wires 8 are inserted from both ends to ensure good ohmic contact between their ends and the conductive slurry inside the tube. The 3D-printed port fixing mold 11 for end fixation and electrode lead-out is used to seal and fix both ends of the rubber tubing 10 to prevent slurry leakage and ensure stable electrode position. Finally, the encapsulated component is placed in a constant temperature drying oven at 80°C for 2 hours to heat and cure. After the PDMS is completely cross-linked and cured, the biomimetic self-sensing tendon 2 with self-sensing function is obtained.

[0048] Those skilled in the art should connect all electrical components and their compatible power supplies in this case via wires, and should select appropriate controllers according to actual conditions to meet control requirements. The specific connection and control sequence should refer to the working principle described below, where the electrical components are connected in sequence. The detailed connection methods are well-known technologies in the field. The following mainly introduces the working principle and process of the self-sensing tendon and drive end-contact end dual cooperative force sensing system, and will not explain the electrical control.

[0049] like Figure 1 As shown, in the embodiment of the present invention, the two ends of the bionic self-sensing tendon 2 are respectively fixed between the output end of the linear servo motor 5 of the line-driven dexterous hand drive system and the input end of the dexterous hand actuator, forming a force transmission path. Simultaneously, a flexible pressure-sensitive sensing unit is fitted and installed at the contact position at the end of the actuator to directly sense external fingertip pressure. The leads of both types of sensors are connected to the main control signal acquisition module, forming a complete sensing loop. After system power-on initialization, the tension zero point and pressure reference calibration process is automatically completed. Initial drift and environmental interference signals are effectively filtered out through a software filtering algorithm, thereby establishing a stable and reliable signal baseline. When the linear servo motor 5 is started, the tension causes the bionic self-sensing tendon 2 to undergo axial tensile deformation. The conductive network of the nano-conductive composite material composed of multi-walled carbon nanotubes (MWCNTs) and polydimethylsiloxane (PDMS) inside the tendon is subsequently reconstructed.

[0050] This causes a change in the overall resistance value. The resistance signal changes linearly with the applied tension (with a piecewise linearity R² greater than 0.99 within the 0-20 N range). After analog-to-digital conversion and calculation using a preset calibration algorithm, the tension value within the 0–20 N range can be output accurately and in real time. Meanwhile, the pressure generated when the end effector of the dexterous hand comes into contact with an external object is converted into a capacitive signal by the flexible pressure-sensitive sensing unit and synchronously transmitted to the controller. This pressure sensor is designed based on the principle of ionogel capacitance, featuring a fast response time of less than 100ms and excellent linearity (0-10 N). (0.997); the dual-cooperative force sensing system at the drive end and contact end in the controller processes these two signals in real time.

[0051] The specific strategy for sensing the combined tension and compression forces is as follows:

[0052] In this invention, the collaborative force sensing employs a dual-signal collaborative control strategy of driving-end tension and end-effector pressure. When the dexterous hand attempts to grasp an object after acquiring it through the vision module, and the contact pressure signal fed back by the flexible pressure-sensitive sensing unit is invalid and unstable, the force control accuracy and the smoothness of the execution action are significantly improved through dual-end sensor fusion and complementary verification. The strategy flowchart is attached. Figure 8 As shown, after system initialization, data from the bionic self-sensing tendon and fingertip pressure sensor are collected to determine whether contact has occurred. Once contact is confirmed, the degree of tendon stretch is determined by the servo motor rotation angle, and the dexterous hand posture is reconstructed based on kinematic analysis, as detailed below:

[0053] Forward kinematic analysis of dexterous hand fingers typically employs the DH parameter method to solve for the fingertip pose equations. This involves establishing a joint coordinate system to obtain a DH parameter table that describes the relationships between phalanges, thereby constructing a transformation matrix between adjacent phalanges. Finally, the positional relationship of the fingertip coordinates relative to the base coordinates can be determined. This example uses the index finger and thumb as representatives of the joint coordinate system for forward kinematic analysis.

[0054] As attached Figure 6 As shown, a simplified joint coordinate system is established based on the structure of the index finger. The DH parameters are determined using relevant parameters of the index finger in a dexterous hand, as shown in Table 1. This indicates the length of the connecting rod, specifically the length of the common perpendicular between the axes of two adjacent joints. This indicates two connecting rods along two common perpendiculars. and The length between them is such that the finger joints are all located on the same plane. =0; It is expressed as the torsional angle of the link, that is, the angle between the axes of the two joints; This is represented as the linkage angle, which is the angle of rotation of the three joint angles. When the joint rotates, It changes accordingly.

[0055] Table 1. Index finger DH parameters

[0056]

[0057] Based on the requirements of the DH parameters, establish the transformation matrix between two adjacent coordinate systems, i.e. and Coordinate system transformation matrix As shown in formula (1):

[0058] (1)

[0059] The coordinate transformation matrix between the links of the fingers can be obtained from formula (1), and the homogeneous coordinates of each coordinate system can be further determined, as shown in formulas (2), (3), and (4):

[0060] (2)

[0061] (3)

[0062] (4)

[0063] Combining Table (1) with the above transformation matrix, we can obtain the transformation matrix of the index finger base coordinate system relative to the fingertip coordinate system, as shown in Formula (5).

[0064] (5)

[0065] In the formula, = = , = = , = , = ), = , = ); , , Representing 1, 2, and 3 respectively; the third-order matrix in the upper left corner of the final matrix in formula (5) This is the finger pose change matrix, column 4, number 3. 1 matrix Let be the finger position change matrix, that is, the spatial position of the fingertip relative to the base coordinate system; therefore, the coordinate matrix of the index fingertip position and the motion angle of each joint is shown in formula (6):

[0066] (6)

[0067] According to formula (6), once the joint angle of the index finger and the length of the connecting rod are determined, the position of the fingertip of the index finger is uniquely determined by formula (6).

[0068] As attached Figure 7 As shown, a simplified diagram of the thumb coordinate system is established using the DH parameter method, and the detailed parameters of each link are shown in Table (2).

[0069] Table 2 Thumb DH Parameters

[0070]

[0071] The results of the secondary coordinate transformation of each coordinate system are calculated based on the established coordinate system, as shown in formulas (7), (8), and (9).

[0072] (7)

[0073] (8)

[0074] (9)

[0075] The transformation matrix between the thumb tip and the base coordinates is shown in formula (10):

[0076] (10)

[0077] In formula (10), the analytical expressions for m and n are shown in formula (11):

[0078] (11)

[0079] Therefore, the coordinate matrix of the thumb tip position and the movement angles of each joint is shown in formula (12):

[0080] (12)

[0081] Therefore, after determining the angles of each joint of the thumb and the length of the connecting rod, the position of the thumb tip is uniquely determined by formula (12). Since a dexterous hand is driven by underactuated tendons, the tendon stretching length has a mapping relationship with the rotation angle of each joint, as follows:

[0082] As attached Figure 9 As shown, a simplified analysis diagram of the index finger movement in a bionic line-driven dexterous hand is presented. The distance between the tendon line between the fingertip node and the servo motor connection node is... and The sum of these factors means that the tendinous lines will move when the dexterous hand flexes and extends. As the distance contracts, joints 1, 2, and 3 rotate respectively. , and As shown in the upper lower half, actually The distance is The total change is given by the cosine theorem of the triangle in the figure, as shown in formula (13):

[0083] (13)

[0084] Therefore, the amount of change in the internal tendons of the fingers The analytical expression is shown in formula (14):

[0085] (2-14)

[0086] In the formula , 1, 2, 3; Further calculate the angles between the tendinous line and the sides of each link. and As shown in formula (15):

[0087] (15)

[0088] Determining the tendon stretch length using the known servo angle The angle changes of each joint are then determined, and the posture of each finger joint is reconstructed through inverse kinematics. Simultaneously, the host computer processes the data from the bionic self-sensing tendon and fingertip sensors, performing signal noise reduction and smoothing to remove noise and interference. Then, the original electrical signals are converted into tendon tension and fingertip pressure data using a pre-calibrated mapping model, and time-stamp alignment and interpolation resampling are used to achieve dual-end data synchronization. After data processing, tendon tension and fingertip pressure are complementaryly fused using an adaptive weighting method to improve data robustness and reliability under conditions of poor or no fingertip contact. The static analysis of tendon tension mapping to fingertip pressure is detailed below:

[0089] As attached Figure 10 As shown, taking the index finger as an example, a force diagram analysis is performed. When the fingertip is subjected to an external load... At this time, since the phalanges are in equilibrium, formula (16) is obtained based on the force equilibrium analysis:

[0090] (16)

[0091] In formula (16), This represents the tension generated when the driver stretches the tendon line. and fingertip load The normal and tangential components of the force at the point of contact. Simultaneously, restoring torsion springs are embedded in each joint of the dexterous hand's fingers, generating a torque of... = , For the torsional stiffness of the torsion spring, The angle of twist is given. The final relationship between the fingertip load and the tendon tension is shown in formula (17):

[0092] (17)

[0093] As can be seen from formula (17), there is a mapping relationship between the positive pressure at the fingertips and the tension on the tendon line at this time, which can provide a theoretical basis for the coordination of force perception at both ends.

[0094] After data fusion, the load at the fingertip is obtained, and the finger driving force is adjusted according to the desired load. Simultaneously, throughout the entire operation, the system continuously monitors the tension value of the bionic self-sensing tendon 2 for safety. Once the tension exceeds the preset 20 N safety threshold, the system immediately activates the overload protection mechanism, reducing the output power of the drive unit or directly cutting off the power source, thereby effectively protecting the tendon body, transmission structure, and actuator from mechanical damage. The advantage of this implementation is that the bionic self-sensing tendon 2 integrates driving and sensing functions, eliminating the need for an additional independent force sensor in the system, greatly simplifying system integration. Its unique dual-end collaborative sensing architecture allows it to maintain continuous and reliable sensing even under complex conditions such as irregular contact, dynamic impact, or high-speed movement. Therefore, this invention can be widely applied to various tendon-driven systems such as rehabilitation exoskeletons, flexible joints, and bionic robotic arms, significantly improving the system's force control stability, operational safety, and adaptability to unstructured environments.

[0095] Example

[0096] To verify the grasping ability of the dexterous hand, and referring to the common human hand grasping classification system, appropriate grasping gestures were used to grasp nine common objects of different stiffness and shapes in daily life. The main grasping gestures were three primary human hand grasping gestures: interfinal resistance (two-finger pinching, three-finger grasping), lateral resistance (lateral finger pinching), and palm resistance (five-finger enveloping grasping). The results fully demonstrate that the dexterous hand designed in this paper has good stability and grasping flexibility, enabling diverse grasping of different objects. Grasping tests are as follows: Figures 11 to 14 As shown.

[0097] Pinch with two fingers: as Figure 11 As shown, dexterous hands typically use a two-finger pinching method to grasp small, delicate objects, similar to human hands. They use the index finger and thumb to perform fingertip counter-gripping, achieving a stable grip on small, delicate objects. Figure 11 (a) A dexterous hand uses two fingers to pinch a tangerine and achieves a stable grip; Figure 11As shown in the three figures in (b), the dexterous hand first uses two fingers to firmly grasp the tweezers, then places the chip in the center of the tweezers, increases the grasping force to firmly grasp the chip through the tweezers, and finally places it on the PCB board. Throughout the process, by adjusting the appropriate grasping force of the dexterous hand's index finger and thumb, it is possible to achieve stable grasping of small and delicate objects.

[0098] Three-finger grasp: such as Figure 12 As shown, dexterous hands typically use a three-finger grasping method to grasp regular or irregular objects of moderate size. In addition to the two-finger pinching method, the middle finger is added to increase the grasping space and improve the grasping force and stability. Figure 12 (a) The dexterous hand can grasp an orange with three fingers and maintain stability. Even when the position of the dexterous hand is changed or shaken, the dexterous hand can still grasp it stably, indicating that the dexterous hand has good grasping performance for spherical objects. Figure 12 (b) The dexterous hand can grasp a rectangular milk carton stably using three fingers, and even when the position of the dexterous hand is changed or shaken, it can still grasp it stably, proving that the dexterous hand has good grasping performance for rectangular objects; similarly Figure 12 (c) The dexterous hand grasps a disposable paper cup with three fingers. Even if the position of the dexterous hand is changed or shaken, the dexterous hand can still grasp it stably, which shows that the dexterous hand has good grasping performance for cylindrical objects.

[0099] Pinching with the side of the fingers: such as Figure 13 As shown in (a) and (b), to verify the good finger-side grasping performance of the dexterous hand, tests were conducted by grasping a small servo motor and a tangerine. It can be seen that the dexterous hand has excellent grasping performance for flat or small objects using finger-side pinching.

[0100] Five-finger pericardial grasping: such as Figure 14 As shown, dexterous hands typically use a five-finger enveloping grasping method to grasp large, regular or irregular objects. The five fingers work together to adaptively conform to and wrap around the object, maximizing the dexterous hand's workspace and gripping performance. Figure 14 (a) The dexterous hand can grasp a banana and maintain stability. Even if the position of the dexterous hand is changed or shaken, the dexterous hand can still grasp it stably, which shows that the dexterous hand has good grasping performance for irregular long objects. Figure 14 (b) The dexterous hand can grasp a cola can and maintain stability. Even when the position of the dexterous hand is changed or shaken, the dexterous hand can still grasp it stably, which shows that the dexterous hand has good gripping performance on cylindrical objects with high rigidity. Figure 14 (c) The dexterous hand can grasp a full load of about 500 g of mineral water and maintain stability. Even when the position of the dexterous hand is changed or shaken, the dexterous hand can still grasp it stably, which shows that the dexterous hand has good grasping performance for larger objects as well.

[0101] In summary, the dexterous hand demonstrated good grasping performance and reliability for nine common objects of different stiffness and size through two-finger pinching, three-finger grasping, finger-side pinching, and five-finger all-around grasping.

[0102] like Figure 15 As shown, taking the three-finger grasping of an orange as an example, the specific grasping action is as follows: First, the host computer detects the orange through the vision module and obtains its precise coordinates, shape, size, and other information. Based on the object being inspected, the host computer provides a suitable pre-grasping gesture and pre-grasping force and feeds it back to the dexterous hand control system via data packets. Second, the host computer sends motion commands to the robotic arm via TCP communication, plans the motion path based on the orange's pose information, and controls the robotic arm to move the dexterous hand to the predetermined grasping pose of the orange. Third, the dexterous hand control system parses the received data packets to obtain the pre-grasping gesture (the orange is a sphere, so a three-finger grasping gesture is used) and the pre-grasping force (the orange is a flexible, regular object, so the fingertip grasping force is set to 2). (N, to avoid crushing damage), drive the thumb lateral joint and the four finger flexion and extension joints to adjust to the initial grasping posture, and the bionic tendon tension sensor and fingertip pressure sensor enter the real-time sensing ready state; in the fourth step, the dexterous hand begins to grasp according to the pre-grasping gesture, and at the same time, the fingertip pressure sensor and tension sensor work together to sense and control to ensure stable grasping of the orange; in the fifth step, the dexterous hand completes stable grasping; finally, adjust the posture of the robotic arm and verify the grasping stability of the dexterous hand system.

[0103] like Figure 16 As shown, taking the four-finger grasping of a cola can as an example, the specific grasping action is as follows: First, the host computer detects the cola can through the vision module and obtains its precise coordinates, shape, size, and other information. Based on the object being inspected, the host computer provides a suitable pre-grasping gesture and pre-grasping force and feeds it back to the dexterous hand control system via data packets. Second, the host computer sends motion commands to the robotic arm via TCP communication, plans the motion path based on the cola can's pose information, and controls the robotic arm to move the dexterous hand to the predetermined grasping pose of the cola can. Third, the dexterous hand control system parses the received data packets to obtain the pre-grasping gesture (the cola can is a cylinder, using a five-finger enveloping grasp) and pre-grasping force (the cola can is a rigid, regular object; the grasping force of the four fingertips is set to 2 N, and the thumb to 3 N). (N) Drive the thumb lateral joint and the four finger flexion-extension joints to adjust to the initial grasping posture, and the bionic tendon tension sensor and fingertip pressure sensor enter the real-time sensing ready state; Fourth step, the dexterous hand begins to grasp according to the pre-grasping gesture, and at the same time, the fingertip pressure sensor and tension sensor work together to ensure stable grasping of the cola can; Fifth step, the contact of each finger is judged according to the pressure value fed back by the fingertip. If the fingertip force sensing contact fails, the tension control mode is switched to ensure real-time force control of the object grasp; Sixth step, after the grasping is stable, the posture of the robotic arm is adjusted to further verify the grasping stability of the dexterous hand system.

[0104] like Figure 17 As shown, taking the five-finger grasping of a banana as an example, the specific grasping action is as follows: First, the host computer detects the object as a banana through the vision module and obtains its precise coordinates, shape, size, and other information. Based on the object being detected, the host computer provides a suitable pre-grasping gesture and pre-grasping force and feeds it back to the dexterous hand control system via data packets. Second, the host computer sends motion commands to the robotic arm via TCP communication, plans the motion path based on the banana's pose information, and controls the robotic arm to move the dexterous hand to the predetermined grasping pose. Third, the dexterous hand control system parses the received data packets to obtain the pre-grasping gesture (the banana is a long and thin object, so a five-finger envelope grasping method is used) and the pre-grasping force (the banana is a flexible and irregular object, so a pulling force control mode is used, with a desired pulling force of 6). (N) Drive the thumb lateral joint and the four finger flexion-extension joints to adjust to the initial grasping posture, and the bionic tendon tension sensor and fingertip pressure sensor enter the real-time sensing ready state; Fourth, the dexterous hand begins to grasp according to the pre-grasping gesture, and at the same time, the fingertip pressure sensor and tension sensor work together to sense and control to ensure stable grasping of the banana; Fifth, after the tension drive end is stabilized, the fingertip force pressure is obtained to ensure that the contact pressure is appropriate and to prevent excessive squeezing of the banana and damage to the surface; Sixth, after the grasping is stable, the posture of the robotic arm is adjusted to further verify the grasping stability of the dexterous hand system.

[0105] The results of the grasping experiment of the dexterous hand system are shown in Table 3. To further verify the advantages of dual-coordinated sensing and control of fingertip pressure and tendon tension, we conducted 20 experiments on each of the three different objects using both single-fingert pressure sensing and control and dual-coordinated sensing and control of fingertip pressure and tendon tension, and verified their grasping success rates. The table shows that the grasping success rate with dual-coordinated sensing and control is generally higher than that with single-fingert pressure sensing and control. For grasping large, flexible spheres like oranges, the tendon-type dexterous hand has good adaptability, so the fingertips generally have contact sensing, and both grasping methods achieve a 100% success rate. For rigid cylindrical cola cans, sometimes the fingertips cannot make good contact, causing the fingertip force after PID control to gradually increase, seriously damaging the object and the actuator in the dexterous hand itself. Occasionally, it can cause the grasping posture to shift and the object to fall. However, dual collaborative sensing avoids this problem by detecting the tension data in real time, achieving stable grasping by the dexterous hand. For long and irregular objects like bananas, it is also difficult to ensure that the point of contact between the finger and the object is in contact with the fingertip sensor, making it almost impossible to sense. Therefore, it leads to multiple grasping failures under single fingertip control. By coordinating tension control with fingertip sensing, the grasping success rate is greatly improved. However, in rare cases, it is possible to fail to grasp such an irregular object. This is mainly because the banana is placed flat on the table, and the thumb of a dexterous hand is relatively large, which sometimes makes it difficult to make good contact with the curved outline of the bottom of the banana, resulting in unstable contact at the moment of grasping. However, it is often possible to achieve a stable grasp of the banana by using the other four fingers in coordination.

[0106] Table 3. Success Rate of Grasping in Two Different Modes of the Dexterous Hand System

[0107]

[0108] Experimental results show that the coordinated operation of the tension sensor at the drive end and the pressure sensor at the contact end not only overcomes the limitations of single sensing and improves the ability to grasp irregular objects, but also, combined with the target recognition and pose estimation of the vision module, achieves high-success-rate autonomous grasping of different objects, verifying the system's perception reliability and force-controlled grasping stability, and laying the foundation for practical applications.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A bionic self-sensing tendon for a wire-driven dexterous hand, characterized in that, The device includes an elastic hose made of a flexible elastic matrix, a nano-conductive composite material core layer (9) filled inside the elastic hose, and a port fixing mold (1) fixed at both ends of the elastic hose. The outer end of the port fixing mold (1) has a through channel for a polymer polyester tendon wire (3) and a copper wire (8) at the center. After the polymer polyester tendon wire (3) and the copper wire (8) are introduced, they are fixed at one end by spot welding. The copper wire (8) passes through the port fixing mold (1) and connects with the nano-conductive composite material core layer (9) to form an electrode. The tendon wire (3) is connected to the port fixing mold (1) to transmit tensile force.

2. The biomimetic self-sensing tendon as described in claim 1, characterized in that, The flexible elastic matrix is ​​made of rubber tubing (10) with a modulus similar to that of human tendons. The nano-conductive composite core layer (9) is made of multi-walled carbon nanotubes and PDMS, which has excellent flexible tensile and strain resistance characteristics.

3. The biomimetic self-sensing tendon as described in claim 2, characterized in that, The preparation method of the nano-conductive composite core layer (9) is as follows: Weigh multi-walled carbon nanotubes (MWCNTs) as conductive fillers, prepare component A (main body) and component B (curing agent) of polydimethylsiloxane (PDMS), and n-hexane as a dispersant; add the weighed MWCNTs (preferably 7% by mass) to n-hexane, and form a suspension after oscillation in an ultrasonic oscillator; add PDMS to the suspension, stir with a mechanical stirrer for 120 minutes to make MWCNTs uniformly dispersed in the PDMS matrix, and allow n-hexane to completely evaporate; then add PDMS-B curing agent to the above mixture, and continue stirring for 10 minutes to ensure uniform mixing; finally, put the mixed slurry into a vacuum drying oven and perform 3 complete vacuum cycles to completely remove the air bubbles introduced during the stirring process, and obtain a bubble-free uniform conductive slurry.

4. A bionic dexterous hand based on bionic self-sensing tendons, characterized in that, The device includes four fingers that are independently configured and capable of flexion and extension movements, and a thumb that can perform flexion, extension, and lateral swing movements respectively. The four fingers are driven independently by four linear servos (5), the thumb is controlled by a lateral swing servo (6) to swing laterally, and the linear servos control flexion and extension. The wrist is controlled by a motor to rotate a synchronous belt (4) to bend the wrist of the dexterous hand. The flexion and extension movements of the four fingers and the thumb are all driven by tendon-line transmission. Pressure-sensitive sensors capable of sensing pressure are installed at the fingertips of the four fingers and the thumb. The bionic dexterous hand is equipped with the bionic self-sensing tendon described in any one of claims 1 to 3. The bionic self-sensing tendon replaces a section of tendon line on the back of the dexterous hand and is used to detect the driving force of the tendon line. Combined with the fingertip flexible pressure-sensitive sensor, a dual-sensory feedback is formed to achieve precise control of gripping force.

5. The bionic dexterous hand as described in claim 4, characterized in that, The four fingers and thumb have a built-in reset torsion spring (7) in their rotation joints. The wrist joint between the palm and the forearm is connected by a rotating shaft. The rotating shaft at the wrist joint and the forearm are both fixedly installed with a synchronous wheel that is connected by a synchronous belt (4). The forearm is equipped with a wrist joint drive servo that drives the synchronous wheel to rotate and makes the palm swing back and forth.

6. The control method for the bionic dexterous hand according to claim 5, characterized in that, Includes the following steps: S1. The host computer detects the precise coordinates, shape and size information of the object under inspection through the vision module. Based on the object under inspection, the host computer gives appropriate pre-grabbing gestures and pre-grabbing forces and feeds them back to the dexterous hand control system through data packets. S2. The host computer sends motion commands to the robotic arm via TCP communication, plans the motion path based on the pose information of the object being inspected, and controls the robotic arm to move the dexterous hand to the predetermined grasping pose of the object being inspected. S3, the dexterous hand control system parses the received data packet, obtains the pre-grasping gesture and pre-grasping force, drives the thumb lateral swing joint and the four finger flexion and extension joints to adjust to the initial grasping posture, and the bionic tendon tension sensor and fingertip pressure sensor enter the real-time sensing ready state. S4. The dexterous hand begins to grasp according to the pre-grasping gesture. At the same time, the fingertip pressure sensor and the tension sensor based on the bionic self-sensing tendon work together to sense and control the object being inspected to ensure stable grasping.

7. The control method for the bionic dexterous hand as described in claim 6, characterized in that, In step S4, during the dual-cooperative sensing and control of the fingertip pressure sensor and the bionic tendon-based tension sensor, the dexterous hand posture is first reconstructed through kinematic analysis. Then, the original tension and pressure data are denoised and smoothed using bandpass filtering and outlier removal algorithms, respectively. An adaptive weighted fusion strategy is adopted to adjust the fusion weights of the two sensors in real time. When the contact is good, the complementary advantages of the dual-modal information are fully utilized to improve the sensing accuracy. When the data quality of a single sensor deteriorates due to poor contact, its contribution ratio is automatically reduced and the compensation effect of the other channel is enhanced, outputting highly robust and low-latency fused fingertip pressure data. The fingertip load is then adjusted in real time based on the fingertip pressure data.

8. The control method for the bionic dexterous hand as described in claim 7, characterized in that, In step S4, the forward motion analysis of the dexterous hand fingers usually uses the DH parameter method to solve the fingertip pose equation. By establishing a joint coordinate system, the DH parameter table is obtained to describe the change relationship between the phalanges, thereby constructing the change matrix between adjacent phalanges. Finally, the positional relationship between the end-point coordinates and the base coordinates can be obtained.

9. The control method for the bionic dexterous hand as described in claim 8, characterized in that, In step S4, the joint coordinate system of the four fingers is approximated by forward kinematics analysis. Therefore, taking the index finger as an example, its DH parameters are determined using relevant parameters of the index finger in a dexterous hand. This indicates the length of the connecting rod, specifically the length of the common perpendicular between the axes of two adjacent joints. This indicates two connecting rods along two common perpendiculars. and The length between them is such that the finger joints are all located on the same plane. =0; It is expressed as the torsional angle of the link, that is, the angle between the axes of the two joints; This is represented as the linkage angle, which is the angle of rotation of the three joint angles. When the joint rotates, Consequently, the transformation matrix between two adjacent coordinate systems is established based on the requirements of the DH parameters. and Coordinate system transformation matrix As shown in formula (1): (1) The coordinate transformation matrix between each link of the finger is obtained from formula (1), and the homogeneous coordinates of each coordinate system are further determined as shown in formulas (2), (3), and (4): (2) (3) (4) Combining Table (1) with the above transformation matrix, we obtain the transformation matrix of the index finger base coordinate system relative to the fingertip coordinate system, as shown in Formula (5); (5) In the formula, = = , = = , = , = ), = , = ); , , Representing 1, 2, and 3 respectively; the third-order matrix in the upper left corner of the final matrix in formula (5) This is the finger pose change matrix, column 4, number 3. 1 matrix Let be the finger position change matrix, that is, the spatial position of the fingertip relative to the base coordinate system; therefore, the coordinate matrix of the index fingertip position and the motion angle of each joint is shown in formula (6): (6) According to formula (6), once the joint angle of the index finger and the length of the connecting rod are determined, the position of the fingertip of the index finger is uniquely determined by formula (6); The forward kinematic analysis of the joint coordinate system represented by the thumb is as follows: a simplified diagram of the thumb coordinate system is established using the DH parameter method, and the secondary coordinate transformation results of each coordinate system are calculated based on the established coordinate system as shown in formulas (7), (8), and (9). (7) (8) (9) The transformation matrix between the thumb tip and the base coordinates is shown in formula (10): (10) In formula (10), the analytical expressions for m and n are shown in formula (11): (11) Therefore, the coordinate matrix of the thumb tip position and the movement angles of each joint is shown in formula (12): (12) Therefore, after determining the angles of each joint of the thumb and the length of the connecting rod, the position of the thumb tip is uniquely determined by formula (12).

10. The control method for the bionic dexterous hand as described in claim 9, characterized in that, In step S4, due to the use of underactuated linearly driven dexterous hands, the tendon stretching length has a mapping relationship with the rotation angle of each joint, specifically as follows: the tendon distance between the fingertip node and the servo connection node is... and The sum of these factors means that the tendinous lines will move when the dexterous hand flexes and extends. As the distance contracts, joints 1, 2, and 3 rotate respectively. , and In fact The distance is The total change is given by the cosine theorem of a triangle as shown in formula (13): (13) Therefore, the amount of change in the internal tendons of the fingers The analytical expression is shown in formula (14): (2-14) In the formula , 1, 2, 3; Further calculate the angles between the tendinous line and the sides of each link. and As shown in formula (15): (15) Determining the tendon stretch length using the known servo angle Then, the angle changes of each joint are determined and the posture of each finger joint is restored through inverse kinematics; at the same time, the host computer processes the data of the bionic self-sensing tendon and fingertip sensor, performs signal noise reduction and smoothing, and removes noise and interference; then, the original electrical signal is converted into tendon tension and fingertip pressure data through a pre-calibrated mapping model, and the data synchronization between the two ends is achieved by using timestamp alignment and interpolation resampling. After data processing, tendon tension and fingertip pressure are complementaryly fused using an adaptive weighting method to improve data robustness and reliability under conditions of poor or no fingertip contact; the static analysis of tendon tension mapping to fingertip pressure is as follows: Taking the index finger as an example, a force diagram analysis is performed. When there is an external load at the fingertip... At this time, since the phalanges are in equilibrium, formula (16) is obtained based on the force equilibrium analysis: (16) In formula (16), This represents the tension generated when the driver stretches the tendon line. and fingertip load The normal and tangential components of the force at the point of contact; simultaneously, restoring torsion springs are embedded in each joint of the dexterous hand's fingers, generating a torque of... = , For the torsional stiffness of the torsion spring, The torsion angle is given; the relationship between the fingertip load and the tendon tension is shown in formula (17): (17) As can be seen from formula (17), there is a mapping relationship between the positive pressure at the fingertips and the tension on the tendon line at this time, which provides a theoretical basis for the synergistic perception of force at both ends.

Images