A high degree of freedom three-fingered hand and an adaptive force control method thereof

Abstract

The application discloses a high-degree-of-freedom three-fingered dexterous hand and a self-adaptive force control method thereof. The dexterous hand comprises a central fixed seat and three installed finger assemblies. Each finger assembly is composed of four standardized micro joint modules in series, including a rotating module, a swinging module and a double-bending module. The whole machine has twelve active degrees of freedom. The joint module integrates a motor, a four-stage speed reducer, a control board and double encoders. The output gear is embedded with a segmented stiffness elastomer, which has both transmission accuracy and impact resistance and buffering capacity. The center hole bearing wiring and the pin jack quick plug connection are used between the modules to realize single-joint quick change. The control method calculates the elastomer deformation variable through the angle difference of the double encoders, combines the pre-stored angle-stiffness mapping curve and the motor current to perform torque closed-loop calculation, adopts a position loop and torque loop dynamic weighted fusion strategy, and realizes flexible adaptive grasping and object soft and hard identification without external sensors. The application is suitable for the fields of automation, robots, rehabilitation equipment and precise operation.

Description

Technical Field

[0001] This invention relates to the field of robotic arm transmission control technology, and in particular to a high-degree-of-freedom three-finger dexterous hand and its adaptive force control method. Background Technology

[0002] The three-finger dexterous hand, possessing the minimum number of fingers required for force application, is one of the earliest forms of dexterous hand and has always been an important research direction in this field. A typical dexterous hand requires at least three fingers, therefore the three-finger dexterous hand is more economical and practical than the five-finger dexterous hand, making it particularly suitable for applications such as industrial automation, robot operation, rehabilitation equipment, and precision assembly.

[0003] Currently, most mainstream three-finger dexterous hand products, both domestically and internationally, have between 3 and 9 degrees of freedom, such as Robotiq and OnRobot. These are primarily designed for specific grasping scenarios and can only perform basic grasping and releasing actions. This low degree of freedom results in insufficient dexterity, poor versatility, a limited range of motion, and a simplistic movement strategy, significantly restricting the application scenarios of three-finger dexterous hands. Furthermore, most existing three-finger dexterous hands concentrate the drive components in the central area, leading to a bulky bottom structure and large size. This layout also makes it difficult to expand to higher degrees of freedom, hindering further performance improvements.

[0004] In terms of maintainability, existing dexterous hands are extremely difficult to assemble and repair. When a dexterous hand malfunctions, it usually requires the entire machine to be returned to the factory for replacement, necessitating complex recalibration. While some dexterous hand manufacturers support replacing individual fingers, they all require the dexterous hand to be removed from the robotic arm (i.e., off-machine maintenance), a cumbersome and time-consuming process that severely impacts production line efficiency. Existing dexterous hands generally lack modular quick-release designs; individual joints or fingers cannot be replaced independently, and on-site recalibration is required after replacement, resulting in high maintenance thresholds and long cycles.

[0005] In terms of impact resistance and structural reliability, dexterous hands, as precision mechatronic products, are relatively fragile. In actual use, when subjected to accidental impacts or drops, damage to the reduction gear and gear breakage frequently occur, leading to the dexterous hand being scrapped or requiring lengthy repairs. Existing dexterous hands generally use rigid gear transmissions, lacking effective impact buffering and energy absorption mechanisms, resulting in weak impact resistance and severely limiting their long-term reliable operation under complex working conditions.

[0006] In terms of force sensing and control, existing dexterous hands generally exhibit rigid grasping processes, lacking adaptive and compliant control capabilities. To achieve force sensing, current solutions typically require external tactile sensors or six-dimensional force sensors, which are not only costly in terms of hardware but also involve complex sensor circuitry and are prone to reliability issues at moving joints. Some solutions rely solely on motor current to estimate output torque, neglecting the effects of gearbox transmission efficiency losses and transmission backlash, resulting in significant torque detection errors and hindering precise force-controlled grasping. Furthermore, existing dexterous hands cannot autonomously identify the hardness or softness of the object during grasping, nor can they automatically buffer and decelerate upon contact, easily leading to problems such as rigid impacts damaging the workpiece or motor stalling and overheating.

[0007] Regarding cable routing reliability, dexterous hands have a large range of motion (typically exceeding 180°), making cable routing an extremely challenging issue in the robotics field. Current dexterous hands often employ external cabling bundles, which, due to repeated bending at the joints, easily lead to open or short circuits after a certain number of uses, severely impacting the hand's lifespan and operational reliability. Furthermore, external cabling also affects the hand's aesthetics and compactness.

[0008] In terms of functional expansion, the fingertip structure of existing dexterous hands is usually a fixed design, which cannot be customized or quickly replaced according to actual working conditions. When the application scenario changes (such as needing to be replaced with sensor fingertips, suction cups, tools, etc.), it is often necessary to replace the entire hand or make complex structural modifications, lacking flexible functional expansion capabilities.

[0009] In summary, existing three-finger dexterous hands suffer from numerous technical bottlenecks, including low active degrees of freedom, poor operational dexterity, difficulty in structural expansion, weak maintainability, insufficient impact resistance, reliance on external sensors for force sensing with low accuracy, poor cable routing reliability, and inability to expand fingertip functions. There is an urgent need for a new type of three-finger dexterous hand and its control method that features high degrees of freedom, modularity, impact resistance, adaptive force control, and ease of maintenance. Summary of the Invention

[0010] To address the above challenges, this invention proposes a novel high-degree-of-freedom three-finger dexterity hand and its adaptive force control method. One of them is a high-degree-of-freedom three-finger dexterous hand, which includes a central fixing base, a three-finger assembly, and a main control board and a power board disposed on the central fixing base; The three finger assemblies are mounted on the central fixed base. Each finger assembly includes a rotary joint module, a swing joint module, a first bending joint module, and a second bending joint module connected in series along the direction away from the central fixed base. Each joint module is a standardized micro joint module that integrates a drive motor, a reduction mechanism, a control board, and a position encoder. The output end of the deceleration mechanism is provided with a module output gear. The module output gear includes an outer ring gear, an inner ring output shaft, and an elastic body connected between the outer ring gear and the inner ring output shaft. The elastic body has a deformation of less than a preset threshold under rated torque conditions and undergoes elastic deformation under impact or overload conditions exceeding the rated torque to absorb impact energy. The joint module is equipped with a motor-end encoder and an output-end encoder to form a dual encoder detection architecture. Adjacent joint modules are connected by rocker arm assemblies, and each joint module and rocker arm assembly is connected by a quick-connect method using pins and sockets; The main control board has a built-in degree-of-freedom adaptive scheduling program. Each joint module communicates in series via a distributed bus. After a joint module is removed or added, the current number of effective degrees of freedom is automatically identified and the corresponding motion control mode is switched.

[0011] Furthermore, the joint module also includes a passive side housing, an intermediate housing, and an active side housing; the reduction mechanism is accommodated between the active side housing and the intermediate housing, the control board is accommodated between the passive side housing and the intermediate housing, and the intermediate housing isolates the mechanical transmission area from the electrical control area; The reduction mechanism is a four-stage stacked gear transmission structure, including a motor output gear, a double stacked gear one, a double stacked gear two, a double stacked gear three, and the module output gear; The motor-end encoder is located at the tail of the drive motor, and the output-end encoder is a magnetic encoder located at the tail of the inner ring output shaft. The motor-end encoder and the output-end magnetic encoder are used to collect the motor-side rotation angle and the actual load output angle, respectively.

[0012] Furthermore, the elastomer is made of polyurethane, and the elastomer is connected to the outer ring gear and the inner ring output shaft through a vulcanization process; The elastic body has segmented stiffness characteristics: in the range from zero to rated torque, the equivalent torsional stiffness is 1200 to 1600 N·mm / rad, corresponding to a torsional deformation of no more than 0.3°; in the range from the rated torque to the stall torque, the equivalent torsional stiffness increases to 8000 to 12000 N·mm / rad, with a maximum allowable torsional deformation of 3°. The control board of the joint module has a pre-stored mapping curve between the torsional angle and stiffness of the elastomer. The mapping curve is obtained by fixing the assembled joint module to a torque testing machine, applying a continuous stepped load from zero to the rated torque of the motor stall, synchronously collecting the torsional angle of the output shaft, recording the torsional stiffness value corresponding to each load, fitting and generating a continuous mapping curve, and then burning it into the local storage chip of the joint module in the form of a discrete array. When the control program runs, the corresponding stiffness value is retrieved in real time using linear interpolation.

[0013] Furthermore, the control board cable of the joint module passes through the rotation center hole of the module housing to the PCB connection board built into the rocker arm assembly. A follower bearing is installed in the center hole. The control board cable passes through the inner ring of the follower bearing. The inner ring of the follower bearing rotates synchronously with the cable to eliminate the relative friction between the cable and the housing. The rocker arm assembly includes a short rocker arm assembly and a long rocker arm assembly. The short rocker arm assembly connects adjacent joint modules in the same axis, and the long rocker arm assembly connects joint modules in different axes. When the rotation axis direction of adjacent joint modules changes, a transition assembly is provided between them. The transition assembly contains a transition PCB board, and the transition PCB board is connected to the rocker arm assembly or the joint module by a quick-connect pin and socket. The fingertip of the finger assembly is connected to the fingertip via a fingertip adapter assembly. The fingertip adapter assembly contains an adapter PCB board. The fingertip has a quick-change structure and can be replaced with soft rubber fingertips of different shapes, tactile sensors, suction cups, or tools.

[0014] Furthermore, the three finger components are finger one, finger two, and finger three, with finger one, finger two, and finger three located on opposite sides of the central fixing base, and finger two and finger three arranged symmetrically. The rotation range of the first finger is ±90°, the rotation range of the second and third fingers is 135° each and the angle ranges are symmetrical; the swing range of the swing joint module is ±90° or more; the bending angle of the first bending joint module near the central fixed seat is ±90° or more, and the bending angle of the second bending joint module near the fingertip is -90° to +115°. The finger assembly and the central fixing base are connected via a bottom PCB board using pin holes. The bottom of the finger assembly is provided with a raised plate to provide space for insertion and removal. The three finger assemblies have the same structure, and the dexterous hand does not distinguish between the left and right hands.

[0015] Furthermore, a cover is provided above the central fixing base, and a central platform covered with soft rubber is provided above the cover; the main control board and the power board are installed inside the cover, and the power board and the main control board are connected by heightening studs; an interface board is provided below the central fixing base, and the interface board is provided with communication interfaces and power interfaces on the side and bottom; the central fixing base is provided with a wiring groove connecting the interface board to the power board or the main control board. Each finger component or joint module can be independently disassembled and replaced, with a replacement time of no more than 30 seconds and no need to remove the dexterous hand from the robotic arm; the joint module has been calibrated and has stored inherent parameters at the factory, and the system automatically calls the inherent parameters after replacement, without the need for secondary calibration on site.

[0016] This invention also provides an adaptive force control method for a high-degree-of-freedom three-finger dexterity hand, applied to the aforementioned high-degree-of-freedom three-finger dexterity hand, comprising the following steps: S1. Parameter preloading: Receive the joint target angle command sent by the host computer, and at the same time retrieve the inherent parameters pre-stored in the joint module. The inherent parameters include the reducer backlash calibration value, the elastomer torsional angle-stiffness mapping curve and the reduction mechanism transmission efficiency coefficient. S2. Multi-channel signal synchronous acquisition: Real-time acquisition of the motor-side rotation angle θ of the encoder at the motor end. m The actual load output angle θ of the output magnetic encoder out And the real-time operating current I of the motor bus; S3. Solving for elastic body deformation: According to the formula Δθ=θ m / iB-θ out Calculate the real-time torsional deformation Δθ of the elastic body, where i is the reduction ratio of the reduction mechanism and B is the backlash calibration value of the reducer; S4. Contact State and Working Condition Identification: The working condition is determined based on the value of the elastic body deformation Δθ and the deformation rate. When Δθ increases rapidly, it is determined that the joint has contacted the workpiece and the motor speed reduction buffer strategy is triggered. When Δθ remains stable, it is determined that the stable clamping state has been entered. The soft and hard properties of the contacting object are distinguished according to the deformation rate. S5. Dual-Loop Fusion Control: Position closed-loop control and torque closed-loop control are executed in parallel. The position closed-loop control corrects the positioning deviation in real time based on the target angle, actual motor rotation angle, backlash value, and elastic body deformation. The torque closed-loop control retrieves the corresponding elastic body stiffness parameters based on the real-time elastic body deformation, and performs a composite correction calculation combining the real-time motor current and the transmission efficiency of the reduction mechanism to solve for the actual joint output torque. The position closed-loop output and the torque closed-loop output are fused and adjusted to coordinately control the output speed and output torque of the drive motor, achieving flexible adaptive grasping.

[0017] Furthermore, the specific algorithm for the dual-closed-loop fusion control in step S5 is as follows: Position closed-loop deviation calculation: E p =θ d -(θ m / iB-Δθ), where θ d The target angle of the joint is calculated using a position loop PI control, which outputs the position adjustment amount U. p ; Torque closed-loop solution: T out =K(Δθ)·Δθ·η+f(I), where K(Δθ) is the real-time equivalent stiffness of the elastic body, η is the transmission efficiency of the reduction mechanism, f(I) is the motor current-electromagnetic torque calibration function, and the torque adjustment amount U is output after torque loop calculation. T ; Dual-loop dynamic weighted fusion: Control output U = α·U p +(1-α)·U T α is the adaptive weighting coefficient for the working condition; under no-load motion conditions, α=1, and only the position closed loop works; under stable grasping conditions, α takes the value of 0.4 to 0.6, and the position loop and torque loop are adjusted in coordination; under impact overload conditions, the value of α is automatically reduced and the torque limiting protection is executed first.

[0018] Furthermore, the reducer backlash calibration value in step S1 is obtained through the following virtual zero-point calibration method: The joint module equipped with the rocker arm assembly is fixed on the calibration fixture. The starting torque threshold is set to 10% of the rated torque. The rocker arm is driven to swing to the left extreme position −90° and the right extreme position +90° respectively. Record the encoder angle values ​​at the motor end and the magnetic encoder angle values ​​at the output end corresponding to the left and right extreme points, respectively. The average value of the output magnetic encoder angles at the left and right extreme points is taken as the virtual zero point. Calculate the difference between the product of the motor encoder motion angle and the reduction ratio and the actual angle of the output magnetic encoder. Use this difference as the backlash value B of the reducer gear and pre-store it in the register of the joint module control board. The calibrated joint module can be directly replaced without the need for secondary calibration on site. When the system is running, it automatically calls the backlash value B for real-time backlash compensation.

[0019] Furthermore, in step S4, the method for identifying the soft and hard properties of the contact object is as follows: In the initial stage of the elastic body deformation Δθ increasing from zero, the deformation rate v = d(Δθ) / dt is calculated in real time; The deformation rate v is compared with a preset threshold: when v is greater than the first preset threshold, the contact object is determined to be a hard workpiece; when v is less than the second preset threshold, the contact object is determined to be a soft workpiece. The gripping strategy is adaptively adjusted based on the recognition results: a rapid deceleration and small torque constant force clamping strategy is adopted for hard workpieces; a slow deceleration and gradual force increase clamping strategy is adopted for soft workpieces. Under stable clamping conditions, the deformation Δθ of the elastic body is continuously monitored. If Δθ changes abruptly, it is determined that the workpiece has slipped or fallen off, triggering a re-grip process or an alarm signal.

[0020] This invention features three identical finger components evenly mounted on a central fixed base. Each finger is composed of four standardized micro-joint modules connected in series for rotation, swing, and double bending, achieving twelve active degrees of freedom. Combined with a four-level layered arrangement ("bottom layer rotation - middle layer swing - top layer double bending"), it effectively avoids interference from multiple joints and fingers, allowing the dexterous hand to operate over the upper hemisphere. This enables specialized dexterous movements such as reverse bending and lateral finger opposition, significantly reducing the need for coordinated robotic arm movements. The joint modules integrate motors, reduction mechanisms, control boards, and dual encoders, achieving high standardization and reusability, reducing manufacturing costs and improving product consistency. The module output gears incorporate elastomers, ensuring transmission accuracy under rated operating conditions. During impact overload, the elastic deformation absorbs energy to protect the internal mechanisms, providing both impact resistance and flexible adaptability. The dual encoder architecture ensures accurate acquisition of the output angle even after the elastomer deforms. The modules are connected by quick-connect pins and sockets, and with the CAN bus adaptive scheduling program, the joint modules can be plugged and played and the degrees of freedom can be increased or decreased as needed. The replacement time for a single joint is no more than 30 seconds and no need to get off the machine, which significantly improves the maintainability and scene adaptability of the dexterous hand.

[0021] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0022] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0023] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 A three-dimensional schematic diagram of a highly flexible three-finger dexterous hand state; Figure 2 This is a three-dimensional schematic diagram of the two states of the high-degree-of-freedom three-finger dexterous hand of the present invention; Figure 3 This is a schematic diagram of the range of motion of the three-finger dexterous hand according to the present invention; Figure 4 The three-finger dexterous hand explosion of the present invention Figure 1 ; Figure 5 The three-finger dexterous hand explosion of the present invention Figure 2 ; Figure 6 This is a three-dimensional view of the fingers of the three-finger dexterous hand of the present invention; Figure 7This is an exploded view of the fingers of the three-finger dexterous hand of the present invention; Figure 8 This is a perspective view of the joint module and rocker arm of the three-finger dexterous hand of the present invention. Figure 9 Another perspective view of the joint module and rocker arm structure of the three-finger dexterous hand of the present invention; Figure 10 This is a perspective view of the joint module structure of the three-finger dexterous hand of the present invention; Figure 11 Another perspective view of the joint module structure of the three-finger dexterous hand of the present invention; Figure 12 This is a schematic diagram of the joint module calibration method for the three-finger dexterous hand of the present invention; Figure 13 This is a perspective view of the central fixing base structure of the three-finger dexterous hand of the present invention; Figure 14 A three-dimensional view of the fingertips of the three-finger dexterous hand of the present invention; Figure 15 This is a perspective view of the module output gear structure of the three-finger dexterous hand of the present invention; Figure 16 This is a plan view of the output gear structure of the three-finger dexterous hand module of the present invention; Figure 17 A perspective view of the three-finger dexterity hand of the present invention, showing the fingers bent in the opposite direction at 90 degrees. Figure 18 Another perspective view of the three-finger dexterity hand of the present invention, with the fingers bent 90 degrees in the opposite direction; Figure 19 This is a three-dimensional view of the three-finger dexterity hand at a 90-degree angle from the side, according to the present invention; Figure 20 This is a perspective view of the three-finger dexterity hand of the present invention, viewed from a 90-degree angle with the fingers facing each other. Figure 21 The diagram showing the mapping between the torsional angle and stiffness of the elastic body used in the three-finger dexterous hand of this invention; Figure 22 This is a block diagram of an adaptive force control method for a high-degree-of-freedom three-finger dexterity hand according to the present invention.

[0024] In the diagram: 1. Finger 1, 2. Finger 2, 3. Finger 3, 4. Middle section, R1. Rotation axis, R2. Swing axis, R3. Bending axis 1, R4. Bending axis 2, 5. Central fixed seat, 6. Central platform, 7. Cover, 8. Extension plate 1, 9. Extension plate 2, 10. Extension plate 3, 11. Main control board, 12. Power board, 13. Interface board, 14. Joint module, 15. Short rocker arm assembly, 16. T-shaped adapter assembly, 17. Planar adapter assembly. 18. Long rocker arm assembly, 19. Fingertip adapter assembly, 20. Fingertip, 21. Long rocker arm PCB board, 22. Joint module control board, 23. Cable, 24. Bottom connection PCB board, 25. Bearing, 26. Passive side housing, 27. Middle housing, 28. Active side housing, 29. Motor output gear, 30. Double stacked gear one, 31. Double stacked gear two, 32. Double stacked gear three, 33. Module output gear, 34. Inner ring, 35. Outer ring, 36. Elastomer. Detailed Implementation

[0025] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the scope of the patent application is not limited or defined by these embodiments. All modifications, equivalents, and substitutions to the embodiments are included within the scope of the claims. In the following description relating to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims. Those skilled in the art will understand the specific meaning of the above terms in this application according to the specific circumstances.

[0026] Example 1: This embodiment provides an overall structure for a highly flexible three-finger dexterous hand, such as... Figures 1 to 7 As shown.

[0027] The dexterous hand includes a central mounting base 5, a three-finger assembly (finger 1, finger 2, and finger 3), and a main control board 11 and a power board 12 mounted on the central mounting base 5. For example, the three fingers are arranged in a clockwise direction, with finger 2 and finger 3 arranged symmetrically and finger 1 on the opposite side. The three fingers have the same degree of freedom in swinging and bending, but the swinging degree of freedom is limited by the structure of the central mounting base, resulting in different ranges of motion.

[0028] Finger assembly structure: Each finger assembly includes a rotary joint module (corresponding to rotary axis R1), a swing joint module (corresponding to swing axis R2), a first bending joint module (corresponding to bending axis one R3), and a second bending joint module (corresponding to bending axis two R4) connected in series along the direction away from the central fixed base 5. This forms a four-degree-of-freedom layered arrangement structure of "bottom layer rotation - middle layer swing - top layer double bending", with a total of twelve active degrees of freedom for the entire machine. Each joint module 14 is a standardized micro joint module that integrates a brushless hollow cup motor, a four-stage reduction mechanism, a control board, and a position encoder. All motion joint modules have the same model, achieving full reuse.

[0029] Three-finger layout: Three finger components are evenly distributed on the central mounting base 5. In a clockwise direction, finger 1, finger 2, and finger 3 are positioned on opposite sides of the central mounting base 5, with finger 2 and finger 3 arranged symmetrically. The rotation range of finger 1 is ±90° (out of a total of 180°), while the rotation ranges of finger 2 and finger 3 are each 135° and symmetrical (due to structural limitations of the central mounting base, finger 2 and finger 3 are reduced by 45° when rotating inward to avoid interference). The swing range of the swing joint module is ±90° or more; due to the different lengths of the rocker arms, their bending angles differ. The bending angle of the first bending joint module (R3) closer to the central mounting base 5 is ±90° or more; the bending angle of the second bending joint module (R4) closer to the fingertip is -90° to +115°.

[0030] Connection between the finger and the central mounting base: The finger assembly is connected to the central mounting base 5 via a pin-and-socket connection on the bottom connecting PCB board 24; other methods, such as press-type contacts or wire bonding, can also be used. When using the pin-and-socket connection, the finger is inserted along the mounting hole on the central mounting base 5, and then the bottom pin is inserted upwards into the connector on the bottom connecting PCB board 24. Elevation plates (elevation plate 1 8, elevation plate 2 9, elevation plate 3 10) are installed below the finger, and the finger is secured with screws on the back. The elevation plates provide space for inserting and removing the pins and sockets.

[0031] Central Fixing Base Structure: A cover 7 is located above the central fixing base 5, and a central platform 6 covered with soft rubber is located above the cover 7. For larger or heavier objects, a finger-enveloping gripping strategy can be adopted, and the soft rubber covering can greatly enhance the gripping success rate. The main control board 11 and power board 12 are installed inside the cover 7. The power board 12 is connected to the main control board 11 via heightened studs. The power board 12 is also connected to the three bottom connecting PCB boards 24 using a quick-connect structure with pins and sockets. This invention provides multiple hardware interface forms, which can be connected to the side or bottom of the dexterous hand; an interface board 13 is located below the central fixing base 5. The interface board 13 has communication interfaces and power interfaces on the side and bottom. The central fixing base 5 has a wiring channel inside to connect the interface board 13 to the power board 12 or the main control board 11. A wiring hole is provided in the middle of the central fixing base 5, which can be used for vertical connection using physical cables or flexible ribbon cables.

[0032] Bus and Adaptive Scheduling: The main control board 11 has a built-in degree-of-freedom adaptive scheduling program, and each joint module 14 communicates in series via a distributed CAN bus. After a joint module is removed or added, the system automatically identifies the current number of effective degrees of freedom and switches the corresponding motion control mode. For example, after removing a joint module, the system automatically switches from 12-degree-of-freedom mode to 11-degree-of-freedom mode; if an entire finger is removed, it switches to 8-degree-of-freedom two-finger mode.

[0033] Internal components of the finger: such as Figure 4 As shown, each finger assembly comprises 4 joint modules 14, 3 short rocker arms 15, 1 T-shaped adapter 16, 1 planar adapter 17, 1 long rocker arm 18, 1 fingertip adapter 19, and 1 fingertip 20. When the rotation axis direction of adjacent joint modules does not change (e.g., between R3 and R4), the rocker arm assembly is used for direct connection; when the rotation axis direction changes (e.g., between R1 and R2, or between R2 and R3), an adapter assembly (T-shaped adapter 16 or planar adapter 17) is required. The adapter assembly contains an adapter PCB board, which can be quickly docked with the rocker arm assembly or joint module.

[0034] This embodiment increases the active degrees of freedom of the three-finger dexterity hand to at least 12, making it the three-finger dexterity hand with the highest active degrees of freedom currently available on the market. Compared to existing technologies such as Robotiq and OnRobot, which only offer 3-9 degrees of freedom, this invention significantly improves dexterity, enabling fingers to bend in the reverse direction by 90° (…). Figure 11 ), 90° to the side (finger) Figure 12 This invention enables special dexterity movements, with an operating range that almost covers the entire upper hemisphere, exceeding the capabilities of the human hand. All the joint modules in this invention are identical, allowing for adjustment of the swing angle with different rocker arms, and enabling the conversion of degrees of freedom through different adapter components.

[0035] This embodiment employs a four-degree-of-freedom layered structure: "bottom layer rotation - middle layer swing - top layer double bending," effectively avoiding interference issues in multi-joint and multi-finger movements. Any two fingers can interact at multiple points, with one finger acting as the active finger and the other two as auxiliary fingers, enabling various three-finger interaction strategies. The high degree-of-freedom operating range reduces the range of motion of the robotic arm, allowing many operations to be completed within the dexterous hand without the need for a robotic arm.

[0036] In this embodiment, the three fingers have identical structures, allowing for dexterity without distinguishing between the left and right hands, significantly reducing product SKUs and inventory costs. It can grasp objects up to 275mm in size, and its opening and closing range is far greater than similar products.

[0037] This embodiment adopts a modular design, which significantly reduces manufacturing costs. All joint modules are standardized products of the same model, which are highly reusable and conducive to quality control and product consistency assurance.

[0038] Example 2: This embodiment describes in detail the internal structure of the joint module 14 and the elastic body design of the module output gear 33, such as... Figures 8-16 As shown.

[0039] The joint module 14 comprises a passive-side housing 26, an intermediate housing 27, and an active-side housing 28. The reduction gear mechanism (mechanical transmission area) is housed between the active-side housing 28 and the intermediate housing 27, while the control board (electrical control area) is housed between the passive-side housing 26 and the intermediate housing 27. The intermediate housing 27 completely isolates the mechanical transmission area from the electrical control area, preventing lubricating oil in the reducer from contaminating the circuit board. This ensures that mechanical transmission and electrical control do not interfere with each other, improving the long-term stability and service life of the module. One side of the joint module is the active end (output power), and the other side is the passive end (for cable routing), connected by a rocker arm.

[0040] Four-stage reduction mechanism: The reduction mechanism is a four-stage spur gear transmission structure, including motor output gear 29, double spur gear one 30, double spur gear two 31, double spur gear three 32, and module output gear 33. The four gears mesh sequentially: motor output gear 29 meshes with double spur gear one 30, double spur gear one 30 meshes with double spur gear two 31, double spur gear two 31 meshes with double spur gear three 32, and double spur gear three 32 meshes with module output gear 33. A reduction ratio exceeding 400:1 is achieved within a limited space, enabling the joint module to achieve stable high torque output while maintaining miniaturization. A brushless coreless motor is recommended to extend the module's lifespan.

[0041] Dual encoder detection architecture: The position encoder includes a motor-end encoder located at the tail of the drive motor and an output-end magnetic encoder located at the tail of the inner ring output shaft 34. The motor-end encoder is used to acquire the motor-side rotation angle θ. m The output magnetic encoder is used to accurately obtain the actual load output angle θ of the joint. out The two together form a dual encoder detection architecture, which can accurately obtain the joint position and accurately determine the actual output angle even after the elastomer has deformed.

[0042] Module output gear elastomer structure: such as Figure 15 and Figure 16 As shown, the module output gear 33 includes an outer ring gear 35, an inner ring output shaft 34, and an elastomer 36 vulcanized between the outer ring gear 35 and the inner ring output shaft 34. The outer ring gear 35 is a gear structure, and the inner ring output shaft 34 is a cross-shaped output shaft, with the elastomer 36 connected in the middle through a vulcanization process. The elastomer 36 is made of polyurethane (PU), but can also be rubber or a spring-like material, with a Shore hardness of 60A (range 55A to 65A), possessing mechanical properties such as high resilience, fatigue resistance, good low-speed flexibility, and significant impact hardening.

[0043] Segmented stiffness characteristics: Elastomer 36 exhibits segmented stiffness characteristics, showing only slight deformation under the motor's rated torque, and greater stiffness beyond the motor's stall torque. Within the rated torque range (0 to rated torque), the elastomer's equivalent torsional stiffness is 1200–1600 N·mm / rad, corresponding to a torsional deformation ≤0.3°, ensuring transmission accuracy. The elastomer hardly deforms during motor rated torque operation. Within the overload range (>rated torque to stall torque), the elastomer stiffness rapidly increases to 8000–12000 N·mm / rad, with a maximum allowable torsional deformation of 3°. When the gear is subjected to impact or force exceeding the motor's rated torque, the elastomer deforms to absorb energy, protecting the reducer gears and motor. Once the deformation reaches a certain range, the elastomer stiffness rapidly increases, preventing excessive deformation and structural damage.

[0044] Elastic body angle-stiffness mapping curve calibration, such as Figure 21 As shown: The assembled joint module is fixed to a torque testing machine, and a continuous stepped load from zero to the rated torque of the motor stall is applied. The torsional angle of the output shaft is collected synchronously, and the torsional stiffness value corresponding to each load is recorded. A continuous mapping curve is then generated by fitting the data. This curve is burned into the local storage chip of the joint module in the form of a discrete array. When the control program runs, the corresponding stiffness value is retrieved in real time using linear interpolation.

[0045] This embodiment adopts a shell partition isolation design, which completely isolates the reducer lubricating grease from the circuit board, avoiding short circuit faults caused by lubricating oil contaminating the circuit board in traditional integrated designs, and significantly improving the long-term operational reliability of the module.

[0046] This embodiment adopts a four-stage stacked gear structure, achieving a reduction ratio of over 400:1 within a very small volume. Compared with traditional two-stage or three-stage reduction schemes, the output torque is significantly improved within the same volume, meeting the requirements of precision gripping and heavy-duty operations, while also providing a structural basis for the miniaturization of dexterous hands.

[0047] The segmented stiffness design of the elastomer output gear in this embodiment achieves a triple function of "precision-buffering-protection": ensuring transmission accuracy under rated operating conditions (deformation ≤0.3°); absorbing energy through elastomer deformation under impact conditions to protect the reducer gear from damage; and drastically increasing stiffness at overload limits to prevent excessive structural deformation. Compared to traditional rigid gear solutions, the dexterous hand's impact resistance is increased several times, and the damage rate after drops or collisions is significantly reduced.

[0048] This embodiment employs a dual-encoder architecture, solving the industry-wide problem of output angle detection distortion after elastic body deformation. Traditional single-encoder solutions cannot determine the actual output angle after elastic body deformation. This invention directly detects the load-side angle through an output-end magnetic encoder, ensuring that position control accuracy is unaffected by the elastic body.

[0049] This embodiment uses an experimentally calibrated angle-stiffness mapping curve, such as... Figure 21 As shown, this method enables torque detection accuracy to far exceed that of traditional methods that rely solely on motor current estimation, providing a reliable hardware foundation for sensorless force control.

[0050] Example 3: This embodiment describes in detail the wiring method and quick-connect structure of the dexterous hand, such as... Figure 8 and Figure 9 As shown.

[0051] Center hole bearing wiring: The control board cable of joint module 14 is routed through the center hole of the module housing to the PCB connection board built into the rocker arm assembly. A follower bearing 25 is installed in the center hole, and the control board cable (cable 23) passes through the inner ring of the follower bearing 25. When the joint module moves, the inner ring of the bearing rotates synchronously with the cable, completely eliminating relative friction and wear between the cable and the housing. Since the range of motion of the dexterous hand joint module exceeds 180°, the reliability of the cable routing is extremely critical under such a large range of motion. This solution avoids repeated bending of the cable at the joint.

[0052] Built-in PCB board in the rocker arm: The rocker arm assembly contains a PCB connection board. The control board cable of the joint module passes through the central hole and connects to the rocker arm PCB board. The rocker arm assembly includes a short rocker arm assembly 15 and a long rocker arm assembly 18. The short rocker arm connects adjacent joint modules in the same axis, while the long rocker arm connects joint modules in different axes. The cable originates from the control board 22 within the joint module, passes through the rotation center hole of the joint module housing and the PCB slot of the rocker arm, and then reverses through the PCB to connect to the PCB, minimizing cable bends and ensuring the cable's secure connection to the PCB connection board.

[0053] Quick-connect pin and socket configuration: All joint modules, rocker arm assemblies, and adapter assemblies utilize a quick-connect pin and socket connection. Pins extend from the tail of the joint module, and sockets are provided on the rocker arm for direct and quick insertion and removal. This pin and socket connection method is not only convenient and reliable but also withstands higher voltage and current limits. When the rotation axis direction of adjacent joint modules changes, an adapter assembly (T-type adapter assembly 16 or planar adapter assembly 17) is provided between them. The adapter assembly contains an adapter PCB board, which also uses a quick-connect pin and socket connection with the rocker arm assembly or joint module.

[0054] Fingertip Quick-Change Structure: The fingertip of the finger assembly is connected to the fingertip 20 via a fingertip adapter assembly 19, which contains an adapter PCB board. The fingertip 20 is a quick-change structure. The fingertip provided in this embodiment is made of soft rubber material with annular protrusions on the surface to increase contact friction. The fingertip can be customized or replaced according to working conditions: it can be replaced with soft rubber fingertips of different shapes and hardness, various tactile sensors, suction cups, ultrasonic probes, or other tools to expand the functionality of the dexterous hand. Combined with the adaptive capability of the joint module, it is easier to grasp thin plates, films, or fabrics.

[0055] Quick-release finger structure: The three finger components have identical structures. The fingers are connected to the central fixing base 5 via a pin-and-hole connection to the bottom PCB board 24, with a heightening board providing space for insertion and removal. Individual finger components or single joint modules can be independently disassembled and replaced in ≤30 seconds, without needing to remove the dexterous hand from the robotic arm (no need to remove the machine for maintenance). Because the modules are calibrated and have their inherent parameters stored at the factory, the system automatically recalls the parameters after replacement, eliminating the need for on-site recalibration.

[0056] The central hole bearing wiring scheme in this embodiment completely solves the industry pain point of cable bending and breakage in the field of dexterity devices. In traditional solutions, cables are repeatedly bent at the joints, resulting in open circuits or short circuits after a certain number of uses. This invention, through central hole wiring and a follower bearing, ensures that the cable hardly bends or wears during joint movement, and no cable is visible externally. This not only guarantees reliability but also improves aesthetics and increases service life by more than 5 times.

[0057] This embodiment employs a quick-connect pin and socket structure, achieving true "plug-and-play" modular assembly. Compared to traditional wire bonding or screw fixing methods, assembly efficiency is increased several times, and it can withstand higher voltages and currents, ensuring high connection reliability. During maintenance, replacement can be completed simply by removing the fixing screws and pulling out the pins, with maintenance time ≤30 seconds. Compared to the traditional factory return repair mode (which typically takes several days), maintenance efficiency is improved by more than a hundred times.

[0058] This embodiment employs a quick-change fingertip design, giving the dexterous hand exceptional adaptability to various scenarios. The same dexterous hand can be adapted to different working conditions simply by changing the fingertips: soft rubber fingertips for general grasping, suction cups for smooth surfaces, sensor fingertips for precision detection, and tool fingertips for special tasks. This eliminates the need to replace the entire hand, significantly reducing usage costs and scenario switching time.

[0059] This embodiment adopts a quick-disassembly design that eliminates the need to disassemble the machine, reducing production line downtime from hours to minutes, significantly improving the economic benefits for industrial applications.

[0060] Example 4: This embodiment describes in detail the adaptive force control method for a high-degree-of-freedom three-finger dexterity hand, such as... Figure 22 As shown.

[0061] Step S1, Parameter Preloading: When the system is working, it first receives the joint target angle command θ sent by the host computer. d Simultaneously, it retrieves the inherent parameters pre-stored in the internal storage chip of the joint module, including: the reducer backlash calibration value B (obtained through the virtual zero-point calibration method described in Example 5), the elastomer torsional angle-stiffness mapping curve K (Δθ) (obtained through the torque testing machine calibration described in Example 2), and the four-stage reducer transmission efficiency coefficient η.

[0062] Step S2, Multi-channel signal synchronous acquisition: The system synchronously performs multi-channel real-time signal acquisition, acquiring: the motor-side rotation angle θ of the encoder at the motor tail. m The actual load output angle θ of the magnetic encoder at the rear of the output shaft out The real-time operating current I of the motor bus and three other core signals are continuously transmitted back to the computing unit in real time.

[0063] Step S3, Elastic Body Deformation Calculation: Based on the collected dual encoder angle data, the system calculates the real-time torsional deformation of the elastic body using a preset formula: Δθ=θ m / iB-θ out Where i is the reduction ratio of the reduction mechanism (i≥400 in this embodiment), and B is the backlash calibration value of the reducer. This formula accurately separates the actual deformation data caused by the compression and torsion of the elastic body, eliminating the influence of reduction ratio conversion error and gear backlash.

[0064] Step S4, Contact State and Working Condition Identification: The system intelligently identifies the grasping working condition based on the value of the elastic body deformation Δθ and the deformation rate v=d(Δθ) / dt. When Δθ increases rapidly (v is greater than the first preset threshold), it is determined that the dexterous hand joint has contacted the workpiece, and the motor deceleration buffer strategy is immediately triggered to avoid structural damage caused by rigid impact. When Δθ remains stable, it is determined that the system has entered a stable clamping state. The hardness or softness of the contacting object is distinguished by the rate of deformation: a high deformation rate (v > first preset threshold) indicates a hard workpiece, and a fast deceleration and small torque constant force clamping strategy is adopted; a low deformation rate (v < second preset threshold) indicates a soft workpiece, and a slow deceleration and gradual force increase clamping strategy is adopted.

[0065] Step S5, Dual-Loop Fusion Control: After the operating condition is identified, the system initiates dual-path calculations in parallel: adaptive position closed-loop control and precise torque calculation. (1) Calculation of position closed-loop deviation: E p =θ d -(θ m / iB-Δθ), output position adjustment amount U after position loop PI calculation. p This compensates for positioning errors caused by backlash and elastic deformation, ensuring the accuracy of joint movement.

[0066] (2) Closed-loop torque calculation: T out =K(Δθ)·Δθ·η+f(I), where K(Δθ) is the real-time equivalent stiffness of the elastic body (obtained through linear interpolation of the mapping curve), η is the transmission efficiency of the reduction mechanism, and f(I) is the motor current-electromagnetic torque calibration function. The torque adjustment amount U is output after torque loop calculation. T This formula combines the real-time stiffness and deformation of the elastic body with the efficiency of the reducer for composite correction, thereby obtaining the true output torque of the joint and overcoming the large error of traditional torque estimation based solely on motor current.

[0067] (3) Dual closed-loop dynamic weighted fusion: control output U=α·U p +(1-α)·U T Where α is the adaptive weighting coefficient for the operating condition: No-load motion condition: α=1, only position closed-loop operation, ensuring motion accuracy and response speed; Stable gripping condition: α ranges from 0.4 to 0.6, with the position loop and torque loop working together to achieve flexible constant force clamping; Impact overload conditions: Automatically reduce the α value and prioritize torque limiting protection to prevent motor stall, overheating, and burnout, thus achieving structural protection.

[0068] Closed-loop iteration: During the entire machine's operation, the motor encoder, output magnetic encoder, and motor current signal continuously flow back to the signal acquisition end through the feedback channel, performing uninterrupted iterative calculations and dynamically correcting control parameters, forming a high-precision, shock-resistant, and adaptive closed-loop force control system throughout the process. The motor power drives the fingers to complete the gripping action through a four-stage reduction module and a built-in elastomer output gear, achieving flexible adaptive envelope clamping of the target workpiece.

[0069] This method achieves force sensing and force-controlled grasping without external sensors. Traditional solutions rely on external tactile sensors or six-dimensional force sensors for force control, which is costly, has complex circuitry, and is prone to damage. This invention can obtain the elastic body deformation and output torque simply by calculating the difference between dual encoders, without any external sensors. This reduces hardware costs by more than 80% and avoids the reliability issues of sensor circuitry at moving joints.

[0070] This method employs multi-parameter fusion torque calculation (elastic body stiffness + deformation + motor current + reducer efficiency), significantly improving torque detection accuracy compared to the traditional single motor current estimation scheme. Traditional schemes do not consider reducer efficiency losses and elastic body deformation, resulting in torque estimation errors of 20%–30%. This invention, through multi-dimensional fusion calculation, controls the torque detection error to within 5%.

[0071] This method employs a dual-closed-loop dynamic weighted fusion strategy, enabling the dexterous hand to respond optimally under various working conditions: rapid and accurate positioning under no-load conditions, flexible constant-force clamping during grasping, and automatic protection during impact. Compared to traditional single-position loop or torque loop control, its adaptability is significantly improved, avoiding workpiece damage or motor stalling and burnout caused by rigid grasping.

[0072] This method has the ability to recognize the hardness or softness of objects, enabling dexterous hands to adaptively grasp various materials (hard metal parts, soft rubber parts, thin sheets and films, fabrics, etc.) without the need for manual preset grasping parameters, thus significantly improving the level of intelligence.

[0073] Example 5: This embodiment details the virtual zero-point calibration method for the joint module and the quick-release maintenance process based on the calibration, such as... Figure 12 As shown.

[0074] Virtual zero-point calibration method: Step 1: Fix the joint module 14 with the rocker arm assembly onto the special calibration fixture, ensuring that the module axis is coaxial with the rotation axis of the fixture.

[0075] Step 2: Using 10% of the rated torque as the starting torque threshold (i.e., the motor drives at extremely low speed and with minimal torque), drive the rocker arm to swing to the left limit position -90° and the right limit position +90° respectively. The purpose of using a low starting torque is to avoid the influence of gear backlash on the determination of the limit point and to ensure that the gears are engaged on one side.

[0076] Step 3: Record the encoder angle values ​​(θ) at the motor end corresponding to the left and right extreme points respectively. mL θ mR ) and the output magnetic encoder angle value (θ) outL θ outR ).

[0077] Step 4: Take the average of the output magnetic encoder angles at the left and right extreme points as the virtual zero point: θ zero =(θ outL +θ outR ) / 2.

[0078] Step 5: Calculate the difference between the product of the motor encoder's motion angle and the reduction ratio and the actual angle of the output magnetic encoder. Use this difference as the backlash value B of the reducer gear. B = (θ) mR -θ mL ) / i-(θ outR -θ outL ) Step 6: Pre-store the backlash value B in the register of the joint module control board, and at the same time store the virtual zero point, elastic body stiffness curve and other inherent parameters.

[0079] Calibration Results: The calibrated joint modules exhibit excellent product consistency and can be directly replaced without requiring secondary on-site calibration. The difference between the actual limit point and the theoretical limit point is used as the backlash value of the joint module. During equipment operation, the backlash value B is automatically called for real-time backlash compensation, eliminating the transmission backlash generated during the forward and reverse motion of the reducer gears, thus giving the joint module higher motion accuracy and repeatability.

[0080] Quick-release maintenance procedure: When a joint module malfunctions or needs to be replaced, the maintenance procedure is as follows: Step 1: Diagnose the location of the faulty module via CAN bus diagnostics; Step 2: Remove the fixing screws (usually 2 to 4) at both ends of the faulty module; Step 3: Disconnect the pin connections between the faulty module and the adjacent rocker arm / adapter assembly; Step 4: Remove the faulty module and install the new calibrated joint module; Step 5: Insert the pins into the sockets to complete the electrical connection, and tighten the screws; Step Six: After the system is powered on, the main control board automatically recognizes the new module, calls its pre-stored inherent parameters (backlash value, stiffness curve, etc.), the CAN bus automatically reconfigures the number of degrees of freedom, and switches the corresponding motion control mode.

[0081] The entire replacement process takes ≤30 seconds, without needing to remove the dexterous hand from the robotic arm (no need to get off the machine), without needing to recalibrate, and without needing special tools.

[0082] Adaptive scheduling of degrees of freedom: The dexterous hand employs a distributed CAN bus serial architecture, with a built-in adaptive scheduling program on the main control board. After removing or adding joint modules, the system automatically identifies the current number of effective degrees of freedom and sensors via bus communication, and automatically switches the corresponding motion control mode. For example, after removing a bending joint module, the system automatically sets that joint to a fixed state, and the kinematic model is automatically updated; if an entire finger is removed, the system switches from a three-finger 12-DOF mode to a two-finger 8-DOF mode.

[0083] The virtual zero-point calibration method used in this embodiment cleverly utilizes the averaging of left and right limit symmetry to establish the joint zero position without relying on an external high-precision reference. It offers high calibration accuracy and ease of operation. Precise calibration and real-time compensation of the backlash value enable the joint repeatability accuracy to reach within 0.1°, meeting the requirements of precision operation.

[0084] This embodiment adopts a factory pre-calibration + parameter pre-storage scheme, enabling the joint module to have "install and use" characteristics. Traditional solutions require a complex recalibration process after joint replacement (usually requiring specialized equipment and several hours of operation time). This invention moves the calibration to the factory stage, and the system automatically calls up the parameters after on-site replacement, significantly reducing the maintenance threshold.

[0085] This embodiment adopts a 30-second quick-disassembly design that eliminates the need to leave the machine, reducing production line maintenance downtime from several days of traditional return-to-factory repairs to less than 1 minute, which has significant economic value in improving the overall equipment efficiency (OEE) of industrial automation production lines.

[0086] The CAN bus adaptive scheduling in this embodiment enables the dexterous hand to have true "reconfigurability". Users can add or remove degrees of freedom according to actual needs (such as simplifying 12 degrees of freedom to 9 degrees of freedom to reduce costs, or increasing to 15 degrees of freedom to improve dexterity), and the system will automatically adapt without modifying the control program or redeveloping it, which greatly improves the product's flexibility and life cycle value.

[0087] In the calibration method of this embodiment, an extremely low starting torque of 10% of the rated torque is used, which effectively avoids the interference of gear elastic deformation on backlash measurement under large driving force, ensuring the accuracy and repeatability of calibration results, and guaranteeing the consistency between different batches of modules.

[0088] Example 6: This embodiment describes the working state and degree-of-freedom expansion scheme of dexterous hands in different application scenarios.

[0089] Scenario 1: Precision gripping of hard workpieces. When the dexterous hand grips a metal screw, the elastic deformation Δθ increases rapidly the instant the finger contacts the workpiece. If the deformation rate v exceeds a first preset threshold, the system determines it is a hard workpiece and immediately triggers a motor deceleration buffer strategy to avoid rigid impact. It then enters a stable clamping state, with α set to 0.5. The position loop and torque loop work together to clamp the screw with a constant torque. After the finger completes a 180° rotation and picks up the screw from the side, the finger returns to its upright position to perform the operation; this action requires no robotic arm intervention.

[0090] Scenario 2: Flexible Grasping of Soft Objects. When the dexterous hand grasps a silicone pad, the elastic deformation Δθ increases slowly. If the rate of deformation v is lower than a second preset threshold, the system determines it to be a soft workpiece and adopts a slow deceleration and progressive force-increasing clamping strategy to avoid excessive clamping force that could deform or damage the workpiece. The flexible characteristics of the elastic body, combined with the progressive force-increasing control algorithm, achieve a smooth effect similar to a human hand grasping soft objects.

[0091] Scenario 3: Pinching thin sheet / film objects. Utilizing the adaptive capabilities of the joint module and the customized design of the fingertips (such as thin, soft rubber fingertips), a dexterous hand can stably pinch thin sheet, film, or fabric objects. The micro-deformation capability of the elastomer allows the fingertips to adapt to the surface contours of extremely thin workpieces, and dual closed-loop control ensures precise and controllable clamping force, preventing breakage or slippage.

[0092] Scenario 4: Enveloping Grasp of Large Objects. For large or heavy objects (diameter ≥ 275mm), a finger-enveloping grasping strategy is adopted. The three fingers fully utilize the large angular range of motion of 12 degrees of freedom (finger bending 90° in the opposite direction, finger facing each other 90° laterally, etc.) to envelop the object in the palm. The soft rubber covering structure of the central platform 6 increases the contact area and friction, improving the success rate of grasping.

[0093] Scenario 5: Fingertip Function Expansion. Depending on the actual working conditions, the fingertip 20 can be replaced with: (a) various tactile sensors for contact force detection in precision assembly; (b) a suction cup for adsorption and handling of objects on smooth surfaces; (c) an ultrasonic probe for non-destructive testing of workpieces; and (d) specialized tools (such as screwdriver heads and clamping jaws) to expand the dexterity of the hand. The adapter PCB board within the fingertip adapter assembly 19 can be configured with a conversion circuit to adapt to the signal interfaces of different sensors.

[0094] Degrees of freedom (DOF) expansion scheme: The modular architecture of this invention supports the addition or reduction of DDFs as needed. Reduction scheme: Remove certain joint modules and replace them with fixed connectors. For example, reduce the four DDFs of each finger to three, making the entire device nine DDFs, suitable for simple grasping and releasing scenarios, and reducing costs. Addition scheme: Add joint modules to the tip of each finger, creating a 15-DOF dexterous hand, further improving dexterity. The system automatically identifies the number of modules via the CAN bus, requiring no modification to the control program.

[0095] Ambidextrous: Since the three fingers are structurally identical and symmetrically arranged, the dexterous hand does not distinguish between the left and right hands during production, assembly, and use. The same product can be installed on either the left or right side of the robotic arm, reducing product models and lowering inventory and management costs.

[0096] This embodiment employs a sensorless adaptive gripping method, enabling the dexterous hand to automatically adapt to various workpieces, including hard, soft, thin, and irregularly shaped ones. There is no need to program or change fixtures for each workpiece individually, and the scene switching time is close to zero, which greatly improves the production efficiency of flexible manufacturing production lines.

[0097] This embodiment can have an ultra-large range of motion with at least 12 degrees of freedom (covering the upper hemisphere), enabling dexterous hands to perform complex movements that are impossible for traditional three-finger dexterous hands, such as... Figure 17 and Figure 18 The reverse bend is shown at 90 degrees. Figure 19 and Figure 20 The 90-degree lateral finger-pointing motion shown significantly reduces the need for coordinated robotic arm movements, lowering system complexity and cycle time.

[0098] This embodiment adopts fingertip quick-change + modular degree of freedom expansion, so that a single dexterous hand platform can cover a wide range of scenarios from simple grasping to precision assembly, from industrial handling to medical testing, and the functional expansion cost within the product life cycle is extremely low.

[0099] This embodiment adopts a design that is suitable for both left- and right-handed users, which halves the number of product SKUs, significantly reduces production, warehousing, and logistics costs, and simplifies user procurement and spare parts management.

[0100] This embodiment employs an envelope gripping strategy in conjunction with a soft rubber central platform, enabling the dexterous hand to stably grasp large, heavy, and irregularly shaped workpieces. The size of the object being grasped is ≥275mm, making it far more applicable than similar three-finger dexterous hand products.

[0101] The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0102] To address the issues of low degrees of freedom and insufficient dexterity, this invention utilizes a joint module approach, rationally arranging components such as the motor, reducer, and control board to achieve miniaturization. This allows for the use of different connecting structures between joints, enabling a dexterous hand to move freely without increasing its size. The miniaturized joint module also achieves a large range of motion (≥180°) with forward and reverse drives, a capability unavailable in dexterous hands using linkages and tendon cables. This results in a significantly larger envelope for the dexterous hand, allowing it to grasp larger objects.

[0103] This invention not only enables the reuse of joint modules but also the reuse of fingers. Considering practicality and economy, three identical fingers are arranged on a central circular fixed base. This symmetrical layout gives the dexterous hand a very large operating range and dexterity. The operating range almost covers the entire upper hemisphere, exceeding that of the human hand. For example, fingers can bend 90° in the opposite direction or bend 90° to the side. The high degree of freedom of operation can reduce the range of motion of the robotic arm, allowing many operations to be completed within the dexterous hand without the need for robotic arm movement.

[0104] This invention reuses joint modules, greatly reducing manufacturing costs and increasing reliability. To solve the problem of not being able to replace a single joint module, this invention not only enables quick replacement of a single finger, but also allows replacement of any joint within a finger, making operation extremely convenient. This is because the modules use a quick-connect method with pins and sockets. Since the modules have position sensors, they are uniformly calibrated at the factory and can be directly replaced, increasing their maintainability.

[0105] Since there is backlash between gears in the reducer during forward and reverse motion, this invention proposes a calibration method for the joint module, which compensates for this backlash in the control algorithm, enabling the joint module to have higher motion.

[0106] To address the issue of easy breakage of the dexterous hand, this invention employs a solid cable, PCB board, and pin socket to avoid cable bending and achieve rapid connection. To reduce cable wear, bearings are used at the module's output end to minimize cable friction, greatly improving the reliability of the module and the entire hand.

[0107] To address the issues of dexterity's inability to withstand impacts and stiffness in the grasping process, this invention embeds an elastomer within the output gear of the joint module. This not only absorbs the energy from accidental impacts to the dexterity, protecting the dexterity and the internal reducer gears, but also, combined with the elastomer's deformation and dual encoder control algorithm, achieves adaptive force control grasping and obtains the torque magnitude of each joint.

[0108] Existing dexterous hands cannot customize the fingertips. To solve this problem, the fingertips of this invention can be customized according to needs or have tools installed, making it easy to expand the functionality of the dexterous hand.

[0109] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A highly flexible three-finger dexterous hand, characterized in that, Includes a central mounting base, a three-finger assembly, and a main control board and a power board mounted on the central mounting base; The three finger assemblies are mounted on the central fixed base. Each finger assembly includes a rotary joint module, a swing joint module, a first bending joint module, and a second bending joint module connected in series along the direction away from the central fixed base. Each joint module is a standardized micro joint module that integrates a drive motor, a reduction mechanism, a control board, and a position encoder. The output end of the deceleration mechanism is provided with a module output gear. The module output gear includes an outer ring gear, an inner ring output shaft, and an elastic body connected between the outer ring gear and the inner ring output shaft. The elastic body has a deformation of less than a preset threshold under rated torque conditions and undergoes elastic deformation under impact or overload conditions exceeding the rated torque to absorb impact energy. The joint module is equipped with a motor-end encoder and an output-end encoder to form a dual encoder detection architecture. The adjacent joint modules are connected by rocker arm assemblies, and the joint modules and rocker arm assemblies are connected by quick-connect pin and socket connection. The main control board has a built-in degree-of-freedom adaptive scheduling program. Each joint module communicates in series through a distributed bus. After a joint module is removed or added, the current number of effective degrees of freedom is automatically identified and the corresponding motion control mode is switched.

2. A high-degree-of-freedom three-finger dexterous hand according to claim 1, characterized in that, The joint module further includes a passive side housing, an intermediate housing, and an active side housing; the deceleration mechanism is housed between the active side housing and the intermediate housing, the control board is housed between the passive side housing and the intermediate housing, and the intermediate housing isolates the mechanical transmission area from the electrical control area; The reduction mechanism is a four-stage stacked gear transmission structure, including a motor output gear, a double stacked gear one, a double stacked gear two, a double stacked gear three, and the module output gear; The motor-end encoder is located at the tail of the drive motor, and the output-end encoder is a magnetic encoder located at the tail of the inner ring output shaft. The motor-end encoder and the output-end magnetic encoder are used to collect the motor-side rotation angle and the actual load output angle, respectively.

3. A high-degree-of-freedom three-finger dexterous hand according to claim 1, characterized in that, The elastomer is made of polyurethane, and the elastomer is connected to the outer ring gear and the inner ring output shaft through a vulcanization process. The elastic body has segmented stiffness characteristics: in the range from zero to rated torque, the equivalent torsional stiffness is 1200 to 1600 N·mm / rad, corresponding to a torsional deformation of no more than 0.3°; in the range from the rated torque to the stall torque, the equivalent torsional stiffness increases to 8000 to 12000 N·mm / rad, with a maximum allowable torsional deformation of 3°. The control board of the joint module has a pre-stored mapping curve between the torsional angle and stiffness of the elastomer. The mapping curve is obtained by fixing the assembled joint module to a torque testing machine, applying a continuous stepped load from zero to the rated torque of the motor stall, synchronously collecting the torsional angle of the output shaft, recording the torsional stiffness value corresponding to each load, fitting and generating a continuous mapping curve, and then burning it into the local storage chip of the joint module in the form of a discrete array. When the control program runs, the corresponding stiffness value is retrieved in real time using linear interpolation.

4. A high-degree-of-freedom three-finger dexterous hand according to claim 1, characterized in that, The control board cable of the joint module passes through the rotation center hole of the module housing to the PCB connection board built into the rocker arm assembly. A follower bearing is installed in the center hole. The control board cable passes through the inner ring of the follower bearing. The inner ring of the follower bearing rotates synchronously with the cable to eliminate the relative friction between the cable and the housing. The rocker arm assembly includes a short rocker arm assembly and a long rocker arm assembly. The short rocker arm assembly connects adjacent joint modules in the same axis, and the long rocker arm assembly connects joint modules in different axes. When the rotation axis direction of adjacent joint modules changes, a transition component is provided between them. The transition component contains a transition PCB board, and the transition PCB board is connected to the rocker arm assembly or the joint module by a quick-connect pin and socket. The fingertip of the finger assembly is connected to the fingertip via a fingertip adapter assembly. The fingertip adapter assembly contains an adapter PCB board. The fingertip has a quick-change structure and can be replaced with soft rubber fingertips of different shapes, tactile sensors, suction cups, or tools.

5. A high-degree-of-freedom three-finger dexterous hand according to claim 1, characterized in that, The three finger components are finger one, finger two, and finger three. Finger one, finger two, and finger three are located on the opposite side of the central fixing base, and finger two and finger three are arranged symmetrically. The rotation range of finger one is ±90°, and the rotation range of finger two and finger three is 135° each with symmetrical angles. In actual use, the rotation angle can be adjusted or limited according to the actual situation. The swing range of the swing joint module is ±90° or more. The bending angle of the first bending joint module near the central fixed seat is ±90° or more, and the bending angle of the second bending joint module near the fingertip is -90° to +115°. The finger assembly and the central fixing base are connected via a bottom connecting PCB board using a pin-and-hole method. The bottom of the finger assembly is provided with a raised plate to provide space for insertion and removal operations. The three finger assemblies have the same structure, and the dexterous hand does not distinguish between the left and right hands.

6. A high-degree-of-freedom three-finger dexterous hand according to claim 1, characterized in that, A cover is provided above the central mounting base, and a central platform covered with soft rubber is provided above the cover; the main control board and the power board are installed inside the cover, and the power board and the main control board are connected by heightening studs; an interface board is provided below the central mounting base, and the interface board has communication interfaces and power interfaces on the side and bottom; the central mounting base has a wiring groove inside to connect the interface board to the power board or the main control board. Each finger component or joint module can be independently disassembled and replaced, with a replacement time of no more than 30 seconds and no need to remove the dexterous hand from the robotic arm; the joint module has been calibrated and stored with inherent parameters at the factory, and the system automatically calls the inherent parameters after replacement, without the need for secondary calibration on site.

7. An adaptive force control method for a high-degree-of-freedom three-finger dexterity hand, applied to the high-degree-of-freedom three-finger dexterity hand as described in any one of claims 1 to 6, characterized in that, Includes the following steps: S1. Parameter preloading: Receive the joint target angle command sent by the host computer, and at the same time retrieve the inherent parameters pre-stored in the joint module. The inherent parameters include the reducer backlash calibration value, the elastomer torsional angle-stiffness mapping curve and the reduction mechanism transmission efficiency coefficient. S2. Multi-channel signal synchronous acquisition: Real-time acquisition of the motor-side rotation angle θ of the encoder at the motor end. m The actual load output angle θ of the output magnetic encoder out And the real-time operating current I of the motor bus; S3. Solving for elastic body deformation: According to the formula Δθ=θ m / iB-θ out Calculate the real-time torsional deformation Δθ of the elastic body, where i is the reduction ratio of the reduction mechanism and B is the backlash calibration value of the reducer; S4. Contact State and Working Condition Identification: The working condition is determined based on the value of the elastic body deformation Δθ and the deformation rate. When Δθ increases rapidly, it is determined that the joint has contacted the workpiece and the motor speed reduction buffer strategy is triggered. When Δθ remains stable, it is determined that the stable clamping state has been entered. The soft and hard properties of the contacting object are distinguished according to the deformation rate. S5. Dual-loop fusion control: Position closed-loop control and torque closed-loop control are executed in parallel. The position closed-loop control corrects the positioning deviation in real time based on the target angle, the actual rotation angle of the motor, the backlash value, and the elastic body deformation. The torque closed-loop control retrieves the corresponding elastic body stiffness parameters according to the real-time elastic body deformation, and performs a composite correction calculation based on the real-time motor current and the transmission efficiency of the reduction mechanism to solve for the actual output torque of the joint. By integrating and adjusting the position closed-loop output and the torque closed-loop output, the output speed and output torque of the drive motor are controlled in a coordinated manner to achieve flexible adaptive grasping.

8. The adaptive force control method for a high-degree-of-freedom three-finger dexterity hand according to claim 7, characterized in that, The specific algorithm for the dual-closed-loop fusion control in step S5 is as follows: Position closed-loop deviation calculation: E p =θ d (θ) m / iB-Δθ), where θ d The target angle of the joint is calculated using a position loop PI control, which outputs the position adjustment amount U. p ; Torque closed-loop solution: T out =K(Δθ)·Δθ·η+f(I), where K(Δθ) is the real-time equivalent stiffness of the elastic body, η is the transmission efficiency of the reduction mechanism, f(I) is the motor current-electromagnetic torque calibration function, and the torque adjustment amount U is output after torque loop calculation. T ; Dual-loop dynamic weighted fusion: Control output U = α·U p +(1-α)·U T , Where α is the adaptive weighting coefficient for the working condition; under no-load motion condition, α=1, and only the position closed loop works; under stable grasping condition, α takes the value of 0.4 to 0.6, and the position loop and torque loop are adjusted in coordination; under impact overload condition, the value of α is automatically reduced and the torque limiting protection is executed first.

9. The adaptive force control method for a high-degree-of-freedom three-finger dexterity hand according to claim 7, characterized in that, The gearbox backlash calibration value in step S1 is obtained through the following virtual zero-point calibration method: The joint module equipped with the rocker arm assembly is fixed on the calibration fixture. The starting torque threshold is set to 10% of the rated torque. The rocker arm is driven to swing to the left extreme position −90° and the right extreme position +90° respectively. Record the encoder angle values ​​at the motor end and the magnetic encoder angle values ​​at the output end corresponding to the left and right extreme points, respectively. The average value of the output magnetic encoder angles at the left and right extreme points is taken as the virtual zero point. Calculate the difference between the product of the motor encoder motion angle and the reduction ratio and the actual angle of the output magnetic encoder. Use this difference as the backlash value B of the reducer gear and pre-store it in the register of the joint module control board. The calibrated joint module can be directly replaced without the need for secondary calibration on site. When the system is running, it automatically calls the backlash value B for real-time backlash compensation.

10. The adaptive force control method for a high-degree-of-freedom three-finger dexterity hand according to claim 7, characterized in that, In step S4, the method for identifying the soft and hard properties of the contact object is as follows: In the initial stage of the elastic body deformation Δθ increasing from zero, the deformation rate v = d(Δθ) / dt is calculated in real time; The deformation rate v is compared with a preset threshold: when v is greater than the first preset threshold, the contact object is determined to be a hard workpiece. When v is less than the second preset threshold, the contacting object is determined to be a soft workpiece; The gripping strategy is adaptively adjusted based on the recognition results: a rapid deceleration and small torque constant force clamping strategy is adopted for hard workpieces; a slow deceleration and gradual force increase clamping strategy is adopted for soft workpieces. Under stable clamping conditions, the deformation Δθ of the elastic body is continuously monitored. If Δθ changes abruptly, it is determined that the workpiece has slipped or fallen off, triggering a re-grip process or an alarm signal.

Images