Rigidity-adjustable soft manipulator and grabbing control method thereof

Abstract

The invention discloses a rigidity-adjustable soft manipulator and a grabbing control method thereof, and belongs to the technical field of soft robots. The manipulator comprises an action execution assembly composed of a plurality of fingers arranged in the circumferential direction, and the action execution assembly is controlled by a driving mechanism to achieve multi-mode movement. Each finger comprises a sawtooth-shaped framework and at least two driving lines penetrating through the interior of the sawtooth-shaped framework in the length direction. The skeleton is integrally formed by adopting a flexible material, and the inner side and the outer side of the skeleton are provided with hole channels for the driving wire to pass through; the tensioning or loosening of the driving wire is controlled through the driving mechanism, so that the length or the bending degree of the finger skeleton is changed, and the stretching and the rigidity dynamic adjustment of the finger are further realized. The problems that an existing soft manipulator is single in function, and the rigidity and the load capacity are contradictory are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of soft robots, and particularly relates to an adjustable-stiffness soft manipulator and a grasping control method thereof. Background Art

[0002] Traditional rigid manipulators have been widely used in the field of industrial automation. However, their inherent rigid structures have limitations such as low safety and poor environmental adaptability when interacting with human-robot collaboration, fine operations, or unstructured environments. Especially when dealing with objects or organisms of different weights and shapes, the rigid structure of the manipulator makes it difficult to adapt to complex and changeable environments. For example, in the grasping applications of underwater vehicle manipulators and the ends of highly compliant robotic arms, they lack flexibility, safety, reliability, and environmental friendliness, and it is difficult to simultaneously achieve high lightweight, high adaptability, and high compliance.

[0003] In recent years, bionic manipulators based on soft materials have shown unique advantages in the fields of medical rehabilitation, service robots, etc. by mimicking the compliant characteristics of biological muscle tissues. However, existing soft hands still face key technical bottlenecks: First, most of them use pneumatic drive or shape memory alloy drive, which requires relying on external air pumps or continuous current excitation, resulting in a large system volume and limited response speed; Second, although the low-stiffness characteristics of the soft structure can ensure operation safety, it is difficult to achieve high-precision positioning and large-load grasping, and the functional singularity significantly restricts the expansion of its application scenarios. In addition, existing cable-driven manipulators mostly focus on single-degree-of-freedom motion (such as pure bending or pure stretching), lack the cooperative control ability of multi-modal deformation, and the coupling mechanism between the drive cable and the structural stiffness is not clear, and it is impossible to dynamically switch between the compliant form and the rigid support. In dynamic grasping scenarios, existing slip detection methods mostly rely on a single index (such as the mean or variance of tension), which is difficult to distinguish sudden slips, gradual relaxation, and noise interference, resulting in a high misjudgment rate and poor adaptability. How to construct a soft operating device with lightweight, multi-modal motion, stiffness dynamic adjustment, and high-robustness grasping is still a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] Technical Problems to be Solved In order to avoid the deficiencies of the prior art, the present invention provides an adjustable-stiffness soft manipulator and a grasping control method thereof. Through cable-driven antagonistic, cooperative, and composite motion control, multi-modal motion of finger bone telescoping, bending, and composite deformation is realized, breaking through the single-degree-of-freedom limitation; and combined with the dynamic coupling mechanism of the serrated skeleton and the drive cable, the stiffness of the finger bone is enhanced when contracting to adapt to heavy loads, and the stiffness is reduced when extending to ensure compliance. The present invention solves the problems of single function and the contradiction between stiffness and load capacity of existing soft manipulators, and can be applied to various environments on land, water, and air.

[0005] The technical solution of the present invention is: a soft manipulator with adjustable stiffness, comprising an action execution component composed of a plurality of fingers arranged along the circumferential direction, wherein the action execution component is controlled by a driving mechanism to realize multi-modal motion; The driving mechanism comprises a driving motor and a winding reel mounted on its output shaft; The finger comprises a sawtooth frame and at least two driving wires running through the frame along the length direction; the frame is formed in one piece with a flexible material, and holes are provided inside and outside the frame for the driving wires to pass through; one end of the driving wire is fixed to a winding reel, and the other end passes through the hole in the finger frame and is connected to the end of the finger, and a miniature tension sensor is installed at the end of the finger for real-time detection of the tension value of the driving wire; By controlling the tension or relaxation of the driving wire through the driving mechanism, the length or curvature of the finger skeleton changes, thereby achieving the extension and retraction of the finger, dynamic adjustment of stiffness and adjustment of the bending angle.

[0006] A further technical solution of the present invention is: the driving mechanism and the action execution component are connected by a rope guiding mechanism, and the rope guiding mechanism serves as a base for fixing the roots of each finger, and is provided with a plurality of fixed pulleys and guide channels for guiding the path of the driving line and reducing friction; the driving line passes through the finger channel and the guide channel at one end of the rope guiding mechanism in sequence, enters the shell of the rope guiding mechanism, passes through a plurality of fixed pulleys, and then passes through the guide channel at the other end of the rope guiding mechanism to reach the driving mechanism, and is connected to the winding reel.

[0007] A further technical solution of the present invention is: the finger includes a sawtooth structure body and a base installed at both ends of the body, the sawtooth structure is composed of N continuous "V"-shaped structures, wherein the angle between the two side walls of the "V" shape is 90°; the base at one end serves as the root of the finger and is installed on the rope guide mechanism, and the base at the other end serves as the end of the finger, in which micro-tension sensors are arranged in the same number as the driving lines and in one-to-one correspondence.

[0008] A further technical solution of the present invention is: the action execution component includes three fingers evenly distributed along the circumferential direction, and each finger has two channels opened in parallel at the inner / outer serrated groove ends, that is, the channels are arranged in a rectangular array along the length direction in space.

[0009] A soft manipulator grasping control method with adjustable stiffness, the specific steps are as follows: The driving mechanism controls the tension or relaxation of the inner driving wire and the outer driving wire to drive the finger to achieve multi-modal motion, which includes: Synergistic movement: Synchronously tighten or relax the inner and outer drive lines to extend and retract the fingers; Antagonistic movement: tighten one side of the drive line and relax the other side of the drive line to bend the fingers inward or outward; Compound motion: Asynchronously tighten the inner drive line and the outer drive line to simultaneously contract and bend the fingers; Real-time detect the tension values of each drive line and calculate the slip criterion based on the tension signals; Adjust the stiffness of the manipulator or trigger a braking operation according to the comparison result between the slip criterion and the dynamic threshold.

[0010] A further technical solution of the present invention is that the slip criterion is obtained by fusing multi-dimensional features, and the calculation formula is as follows:

[0011] where, represents the weight, , λ represents the sensitivity gain; D ( Ti ) represents the tension distribution entropy, represents the fractional derivative, represents the wavelet ridge energy, is the integral of the wavelet ridge energy; i represents the drive line number; t is the current sampling time; α represents the order of the fractional derivative, α ∈(0, 1); s is the scale parameter, s ∈ , fs represents the sampling frequency, f min and f max represent the lowest concerned frequency and the highest concerned frequency; t' represents the translation parameter; The tension distribution entropy: Quantify the distortion of the tension distribution based on the difference between the fourth-order central moment and the variance of the tension values within the sliding window; The fractional derivative: Capture the slow relaxation or stick-slip oscillation of the tension signal through non-integer order derivatives; The wavelet ridge energy: Extract the transient high-frequency energy in the tension signal to locate the slip event.

[0012] A further technical solution of the present invention is that the dynamic threshold is determined by the 99% quantile of the online kernel density estimation, and the calculation formula is as follows:

[0013] where, Q 99 represents the 99% quantile; the online kernel density f (S) estimation function expression is as follows:

[0014] where, Denotes the Yapanenichnikov kernel function; h Denotes the bandwidth; N Denotes the window length of the online kernel density; S ( t ) Denotes the current slip index value; S (k) Denotes the set of historical slip index values { S (1), S (2), ... , S ( N )}, k Denotes the discrete-time index, k = 1, 2, 3, …, N .

[0015] A further technical solution of the present invention is: The comparison between the slip criterion and the dynamic threshold is as follows: When S ( t ) ≤ , the manipulator still maintains the current grasping state; When 2 > S ( t ) > , the manipulator increases the stiffness and enhances the fit to continue grasping; When S ( t ) ≥ 2 , the manipulator brakes, stops working and alarms.

[0016] A further technical solution of the present invention is: The stiffness of the finger is dynamically adjusted by the tension degree of the drive line: Under the cooperative movement, the stiffness increases when the finger contracts and decreases when the finger extends; Under the compound movement, by tightening the drive line asynchronously, the stiffness and the bending angle are adjusted simultaneously.

[0017] A further technical solution of the present invention is: The tension value of the drive line is detected in real time by a micro tension sensor embedded in the finger base and transmitted to the controller for recursive calculation, and the tension mean and variance are updated in real time to avoid repeated operations to improve the response speed.

[0018] Beneficial effects The beneficial effects of the present invention are as follows: An adjustable-stiffness soft manipulator and its grasping control method provided by the present invention break through the limitations of the prior art in multiple dimensions through innovative structural design and intelligent control strategies, and achieve an overall improvement in the performance, adaptability and application scenarios of the soft manipulator. Its core beneficial effects can be summarized as the following aspects: I. High-degree-of-freedom multi-modal motion and dynamic stiffness adjustment ability Traditional soft robotic hands are usually limited by single-degree-of-freedom or fixed-stiffness designs, making it difficult to balance compliance and load capacity. Through the combined control of wire-driven antagonistic, collaborative, and composite motions, the present invention endows the robotic phalanges with multi-modal deformation capabilities: First, the serrated framework structure combined with ultra-high molecular weight polyethylene drive wires can achieve a stretching ratio of up to 900% (from 60 Nmm in the natural state to 6 Nmm in the contracted state) under collaborative motion, significantly expanding the operating range of the robotic hand; Second, by tightening the inner and outer drive wires asynchronously, the phalanges can flexibly bend within a range of ±180°, forming a composite motion in combination with the stretching and contracting actions to precisely adapt to the grasping requirements of different-shaped objects; Third, the dynamic coupling mechanism of stiffness and deformation (enhanced stiffness during contraction and reduced stiffness during elongation) enables the robotic hand to seamlessly switch between compliant grasping and rigid support. For example, when grasping a 5-kg precast concrete block, the phalanges are contracted through collaborative motion to enhance stiffness and ensure load stability; while grasping a 0.1-g electronic chip, a low-stiffness state is maintained to avoid damaging the target object. This stepless stiffness adjustment ability solves the contradiction between "insufficient rigidity" and "excessive flexibility" of traditional soft robotic hands, greatly enhancing the diversity of application scenarios.

[0019] II. High Robustness Grasping and Intelligent Suppression of Slippage in Complex Environments Aiming at the pain point that wire-driven robotic hands are prone to slippage under dynamic loads or external disturbances, the present invention proposes a multi-dimensional slippage detection algorithm based on tension signals, which constructs a self-referential slippage criterion by combining statistical, time-domain, and frequency-domain features S i ( t ) and a dynamic threshold alarm system. Specifically: 1. Tension distribution entropy. Quantify the distortion of the tension distribution through the fourth-order central moment to quickly respond to sudden slippage (such as an instantaneous sharp drop in tension); 2. Fractional derivative. Utilize the long-term memory effect to capture the slow relaxation of the wire (such as progressive slippage caused by friction and wear) to avoid missed detection; 3. Wavelet ridge energy. Focus on high-frequency transient signals (such as frictional vibrations) to early warn of microscopic slippage risks. The three are fused through non-linear weights to cover all types of slippage scenarios, and the alarm threshold (99% quantile) is dynamically adjusted based on online kernel density estimation to effectively suppress interference such as motor noise and temperature drift. In practical applications, when the detected slippage level increases (such as S ( t )≥2 S th ), the robotic hand can immediately trigger emergency braking or adjust stiffness to ensure grasping stability. For example, in the scenario of a drone-mounted robotic hand, when a tree branch sways under the influence of wind, the robotic hand adaptively maintains the contact force by adjusting the stiffness and bending angle in real time to avoid the drone falling due to slippage.

[0020] III. Lightweight Integrated Design and Wide Scenario Adaptability The manipulator of the present invention is integrally formed by 3D printing using lightweight materials such as photocurable resin and medical silicone, and combines a micro tension sensor and a PTFE pulley guiding mechanism to achieve compact structure (the total weight is reduced by more than 40% compared with the pneumatic solution) and high-efficiency drive (the friction coefficient ≤ 0.1). Its modular design supports multi-finger cooperation or independent control and can be flexibly configured on platforms such as the end of a robotic arm, an unmanned aerial vehicle, or an underwater robot.

[0021] IV. Intelligent Control and Efficient Energy Consumption Management Traditional soft manipulators rely on air pumps or continuous current drives, with high energy consumption and lagging response. The present invention adopts a direct drive solution of ultra-high molecular weight polyethylene drive wire and a micro motor, combined with a closed-loop control of tension feedback, significantly reducing energy consumption (the power consumption is reduced by 60% compared with the pneumatic system). The solid lubrication (graphite / molybdenum disulfide) of the drive wire further reduces friction loss and extends the service life. The control algorithm dynamically updates the mean and variance of tension through a recursive formula, avoiding repeated calculations and improving real-time performance (the sampling frequency adapts to the window length N W ). For example, in continuous operations in a logistics warehouse, the manipulator can complete slip warning and adjust the grasping strategy within 0.8 s, and the response speed is more than 3 times higher than that of the traditional solution.

[0022] V. Social and Economic Benefits and Industry Promotion Value The popularization of the present invention will reshape the pattern of automated grasping technology: in the industrial field, its high adaptability can replace manual labor to complete high-risk and high-precision operations (such as nuclear waste treatment and precision assembly), reducing the risk of work-related injuries and production costs; in the field of service robots, the compliant grasping ability helps in scenarios such as elderly care, disability assistance, and medical surgery; in the fields of unmanned aerial vehicles and underwater exploration, it expands the operation range and environmental tolerance of the equipment. It is estimated that its modular design can reduce the manufacturing cost by 30% and at the same time reduce the maintenance cost by 50% (without complex air circuits or hydraulic systems), and has significant market competitiveness.

[0023] In summary, through the deep integration of structural innovation and intelligent control, the present invention has achieved a comprehensive breakthrough in the stiffness adjustment, degrees of freedom of motion, environmental adaptability, and energy efficiency management of the soft manipulator, providing core support for the development of robotics towards a safer, more intelligent, and more ubiquitous direction. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 It is a schematic structural diagram of a soft manipulator with adjustable stiffness in an embodiment of the present invention.

[0024] Figure 2 It is a schematic structural diagram of a finger skeleton in an embodiment of the present invention.

[0025] Figure 3This is a schematic structural diagram of the rope guiding mechanism in the embodiment of the present invention.

[0026] Figure 4 This is the threshold value in the embodiment of the present invention S th Calculation process.

[0027] Figure 5 This is the algorithm flow of the manipulator grasping and sliding detection in the embodiment of the present invention.

[0028] Explanation of reference numerals: 1. Finger, 11. Skeleton, 111. Base, 112. Serrated structure, 113. Channel, 12. Driving wire, 13. Driving wire; 2. Rope guiding mechanism, 21. Installation groove, 22. Guiding channel, 23. Guiding channel; 3. Driving mechanism, 31. Driving motor, 32. Reel, 33. Mounting plate, 34. Connecting column. Detailed implementation manners

[0029] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0030] Based on the problems of the existing soft manipulator such as single function, contradiction between stiffness and load capacity, etc., the present invention proposes an adjustable stiffness soft manipulator including an action execution component composed of a plurality of fingers arranged circumferentially. The action execution component realizes multi-modal motion through the control of a driving mechanism. The driving mechanism includes a driving motor and a reel installed on its output shaft. The finger includes a serrated skeleton and at least two driving wires penetrating through its interior along the length direction. The skeleton is integrally formed by a flexible material, and channels for the driving wires to pass through are provided on both the inner and outer sides. One end of the driving wire is fixed to the reel, and the other end passes through the channel in the finger skeleton and is connected to the finger end, and a micro tension sensor is installed at the finger end for real-time detection of the tension value of the driving wire. By controlling the tightening or loosening of the driving wire by the driving mechanism, the length or bending degree of the finger skeleton is changed, thereby realizing the telescopic, dynamic stiffness adjustment and bending angle adjustment of the finger.

[0031] Specifically, the driving mechanism and the action execution component are connected through a rope guiding mechanism. The rope guiding mechanism serves as a base for fixing the roots of each finger, and is provided with a plurality of fixed pulleys and guiding channels for guiding the path of the driving wire and reducing friction. The driving wire sequentially passes through the finger channel, the guiding channel at one end of the rope guiding mechanism and enters the shell of the rope guiding mechanism, passes through a plurality of fixed pulleys, and then passes through the guiding channel at the other end of the rope guiding mechanism to reach the driving mechanism and is connected to the reel.

[0032] Specifically, the finger includes a sawtooth structure body and a base installed at both ends of the body. The sawtooth structure is composed of N continuous "V"-shaped structures, wherein the angle between the two side walls of the "V" shape is 90°; the base at one end serves as the root of the finger and is installed on the rope guide mechanism, and the base at the other end serves as the end of the finger, in which micro-tension sensors are arranged in the same number as the driving lines and in one-to-one correspondence.

[0033] Specifically, the action execution component includes three fingers evenly distributed along the circumferential direction, and two channels are opened in parallel at the inner / outer serrated groove ends of each finger, that is, the channels are arranged in a rectangular array along the length direction in space.

[0034] The present invention also proposes a soft manipulator grasping control method with adjustable stiffness, and the specific steps are as follows: Step 1: The driving mechanism controls the tension or relaxation of the inner driving wire and the outer driving wire to drive the finger to achieve multi-modal motion. The multi-modal motion includes: Synergistic movement: Synchronously tighten or relax the inner and outer drive lines to extend and retract the fingers; Antagonistic movement: tighten one side of the drive line and relax the other side of the drive line to bend the fingers inward or outward; Compound movement: asynchronous tensioning of the medial and lateral drive lines to cause the fingers to contract and flex simultaneously; Step 2: Detect the tension value of each driving line in real time, and calculate the slip criterion based on the tension signal; Step 3: According to the comparison result between the slip criterion and the dynamic threshold, the stiffness of the manipulator is adjusted or a braking operation is triggered.

[0035] The above technical solution is further described below in conjunction with the accompanying drawings: In one embodiment, referring to Figure 1 As shown, a soft manipulator with adjustable stiffness includes a finger 1, a rope guiding mechanism 2, and a driving mechanism 3.

[0036] Specifically, refer to Figure 1 As shown, the finger 1 includes a skeleton 11, a driving wire 12, and a driving wire 13. The skeleton 11 can be made of light-cured resin, medical-grade silicone, PLA, ABS, TPU, etc., and is integrally formed by 3D printing. The internal pre-buried cable channels (channels 113) on both sides have a diameter of 0.8 to 1.2 mm. The driving wires 12 and 13 are the inner and outer driving wires, respectively. Figure 2As shown, the skeleton 11 includes a base 111, a sawtooth structure 112, and a channel 113. The base 111 is a rectangular parallelepiped, and four micro-tension sensors are embedded and fixed inside, which are respectively connected to the four drive lines. One end of the drive line 12 and the drive line 13 are respectively fixed on the micro-tension sensor to measure the tension value of each drive line, and the other end passes through the channel 113 from top to bottom. The sawtooth structure 112 is composed of N continuous "V"-shaped structures, with an angle of 90°, a thickness of 3mm, and a difference of 30mm between the crest and the trough. Channels 113 are opened from top to bottom on both sides of the skeleton 11. One end of the drive line 12 and the drive line 13 is fixed to the channel at the top of the skeleton 11, and the other end passes through the channel 113 from top to bottom. The drive line 12 and the drive line 13 are made of ultra-high molecular weight polyethylene cables. Solids such as graphite or molybdenum disulfide are used as lubricants between the drive line and the channel.

[0037] Specifically, refer to Figure 3 As shown, the rope guide mechanism 2 is provided with a mounting groove 21, a guide channel 22, and a guide channel 23 on the top. The finger 1 is fixed in the mounting groove 21. The guide channel 22 is provided in the mounting groove 21, and is the input channel of the driving line of each finger. The guide channel 23 is located below the rope guide mechanism 2, and there are only two output channels. The guide channel 22 is the channel of the driving line of each finger, and there are only two guide channels 23, which are the output channels of the driving line 12 and the driving line 13. A plurality of fixed pulleys are provided inside the rope guide mechanism 2. The pulley is made of polytetrafluoroethylene (PTFE), and the surface is polished to reduce the friction coefficient (≤0.1). The pulley diameter is 3 to 5 mm. The cable bending radius is reduced by multi-level guidance to avoid stress concentration. The driving line 12 and the driving line 13 of each finger 1 are respectively input into the rope guide mechanism 2 through the guide channel 22 of the mounting groove 21 where they are located, and then are respectively collected and output through the guide channel 23 after being guided by a plurality of fixed pulleys.

[0038] Specifically, refer to Figure 1 As shown, the driving mechanism 3 includes a driving motor 31, a cable reel 32, a mounting plate 33, and a connecting column 34. The output shaft of the driving motor 31 is fixedly connected to the cable reel 32. The driving motor 31 is fixedly connected to the mounting plate 33. The rope guide mechanism 2 is fixedly connected to the mounting plate through the connecting column 34. One end of the driving wire 12 and the driving wire 13 output from the guide channel 23 are respectively fixedly connected and wound on the cable reel 32. When the driving motor 31 rotates, the cable reel 32 is driven to rotate, thereby driving the driving wire 12 and the driving wire 13, and then pulling the finger 1 to deform.

[0039] In one embodiment, a method for controlling a soft manipulator with adjustable stiffness is provided, wherein the driving wires 12 and 13 are used as driving units of the finger 1, and the driving modes thereof exhibit antagonistic motion, cooperative motion, and compound motion. The specific operations are as follows: Synergistic movement: The synchronous tensioning and relaxation of drive line 12 and drive line 13 promote the telescopic movement of the phalanges. When drive line 12 and drive line 13 are simultaneously tensioned, finger 1 contracts, and its serrated structure 112 reaches a minimum of approximately 6 Nmm. When drive line 12 and drive line 13 are simultaneously relaxed, finger 1 extends from the shortened state back to its original state, and the serrated structure 112 is approximately 60 Nmm in the natural state. Under synergistic movement, the telescopic ratio of the serrated structure 112 can reach 900%.

[0040] Antagonistic movement: Only one of drive line 12 and drive line 13 is tensioned while the other is relaxed to achieve the bending motion. When drive line 12 is tensioned and drive line 13 is relaxed, finger 1 bends inward and the manipulator closes. When drive line 12 is relaxed and drive line 13 is tensioned, finger 1 bends outward and the manipulator opens. Due to its serrated shape and material compliance, it can achieve bending within the range of ±180°, and can further meet the grasping of objects of different sizes within a large range.

[0041] Compound movement: The asynchronous tensioning of drive line 12 and drive line 13 realizes the compound action of telescopic and bending motions. When drive line 12 is tensioned more than drive line 13, finger 1 contracts and bends inward at the same time. When drive line 13 is tensioned more than drive line 12, finger 1 contracts and bends outward at the same time.

[0042] When finger 1 contracts, its stiffness increases, and when it extends, its stiffness decreases. By the stepless and asynchronous telescopic movement of drive line 12 and drive line 13, the stiffness and length of the manipulator can be steplessly adjusted, so as to meet the grasping of objects with different stiffness and different sizes.

[0043] In one embodiment, referring to Figure 4 shown, in the grasping slip detection algorithm of a method for controlling a soft manipulator with adjustable stiffness, in practical applications, the wire-driven manipulator often slides relative to the object when grasping different objects or under environmental disturbances, resulting in unstable and flexible grasping. When the grasped object is not firmly held or is subject to external disturbances, that is, when the soft hand of the present invention slides relative to the target object, the tension value of the drive wire where the soft hand is located ( T i ) changes. Taking T i as the starting point, by detecting the change of T i signal as the basis for sliding discrimination, and then controlling the manipulator. The specific steps are as follows: Step 1: Define three single sliding criteria and define a fused sliding criterion as a dynamic slip index: Step 1.1: Calculate the slip criterion dominated by the tension distribution entropy ( D ( T i )). When grasping an object of unknown mass or when slipping occurs, the tension baseline drifts; when the environmental temperature changes, resulting in changes in the wire stiffness, the tension fluctuates naturally; motor vibration, signal quantization noise, etc. will cause high-frequency jitter in the tension signal, and traditional variance detection is prone to misjudgment. Based on this problem, in this embodiment, the mean value of the tension within the window is calculated in real time based on a sliding window and variance , which are used as local reference benchmarks, and the calculation formula is as follows:

[0044]

[0045] Dynamically update through a recursive formula to avoid repeated calculations:

[0046]

[0047] In the formula, i is the drive wire number, i = 1, 2 are the two drive wires on the inner side of the robotic finger, i = 3, 4 are the two drive wires on the outer side; T represents the tension of the drive wire; T i is the tension value of the i th drive wire; N W is the window length of the tension distribution entropy, which is adaptively adjusted according to the sampling rate , , is the sampling frequency, which is determined by the FFT main frequency of the controller; t is the current sampling time; k is the discrete time index, which is used to traverse each sampling point within the sliding window. Specifically, on the time axis, from the past th sampling point to the current time k = t is the position marker of all tension values in between; is the mean tension, which is the average level of the tension signal and represents the position reference within the sliding window; is the tension variance, which measures the degree to which the tension signal deviates from the mean and reflects the discreteness of the data. In this way, dynamic benchmark tracking is achieved and it is fully adaptive, without the need for manual calibration, and it can adapt to sudden changes in load and environmental interference.

[0048] When slippage occurs, the tension signal not only deviates from the mean value (such as overall relaxation or tension), but also its distribution pattern undergoes a mutation. Instantaneous local slippage causes instantaneous spikes or collapses in tension; continuous slippage makes the tension distribution exhibit skewness or heavy-tail characteristics. The variance (second moment) can only reflect the degree of data dispersion, while the fourth central moment (kurtosis) measures the kurtosis of the tension signal and is more sensitive to the "sharpness" and "tail weight" of the distribution. By the difference between the fourth moment and the variance, the non-Gaussian distribution characteristics caused by slippage are directly quantified, and the spikes / collapses caused by slippage are captured. Therefore, the divergence of the tension distribution based on a sliding window is proposed D ( T i ) as the slippage criterion.

[0049]

[0050] Among them, is the fourth central moment term; is used as a baseline to deduct the influence of normal fluctuations and highlight the contribution of abnormal distributions.

[0051] Step 1.2: Calculate the slippage criterion dominated by the fractional derivative ( ): In the slippage detection of an on-line driven manipulator, the microscopic slippage of the wire (such as stick-slip oscillation on the friction surface) will cause a low-frequency slow change in tension, while the integer-order derivative is insensitive to such signals; motor vibration, electromagnetic interference, etc. will cause the tension signal to contain high-frequency noise, and the traditional derivative operation amplifies the noise. The fractional derivative ( ) enhances the signal-to-noise ratio through the long-term memory effect to capture the gradual process of microscopic slippage, and at the same time has a greater amplification degree for high-frequency components to enhance the anti-high-frequency noise interference, solving the limitations of traditional integer-order derivatives (such as the first-order and second-order) in non-stationary signal processing and microscopic slippage detection.

[0052] Adopt the Grünwald-Letnikov discrete definition of the fractional derivative:

[0053] In the formula, α is the order of the fractional derivative, α ∈(0,1), controlling the memory effect and non-locality of the derivative, and is adjusted in real time dynamically through Hilbert spectrum matching, that is, optimized online according to the working conditions to satisfy , where is the Hilbert transform of the tension signal; is the Gamma function, , calculating the coefficient weight of the fractional derivative to quantify the contribution of historical tension value data; determines each historical sampling point Degree of contribution to the fractional derivative at the current moment; k is the discrete-time index, representing the k th sampling point before the current moment; Δ t is the sampling interval, the discretized time step, t = k Δ t ; is the historical tension value, representing the tension measurement value of the i th drive line at a past moment ; is the truncation order, which controls the memory time M Δ t of the fractional derivative to reduce the computational amount.

[0054] The parameter system of the fractional derivative is co-designed through α , M , Δ t to achieve a balance among the memory effect, computational accuracy, and real-time performance, providing an analysis tool that takes into account both historical accumulation and instantaneous response for slip detection.

[0055] Step 1.3: Calculate the slip criterion dominated by the wavelet ridge energy ( ): Slip events (such as sudden wire relaxation, frictional vibration) will cause short-time transient pulses in the tension signal, and these pulses have non-stationary characteristics in both the time domain and the frequency domain. By introducing wavelet transform and adjusting the scale parameter s and the translation parameter t' , the transient characteristics can be accurately located in the time-frequency domain, the transient energy in a specific frequency band can be extracted, and the sensitivity to micro-slip can be enhanced:

[0056]

[0057] In the formula, is the wavelet ridge energy; t' is the translation parameter, the translation amount of the wavelet on the time axis, representing the center position of the current analysis time window, and locating the occurrence time of the slip event ；s is the scale parameter, s ∈ , fs represents the sampling frequency, f min and f max represent the lowest and highest frequencies of interest, which are set by engineering practice; represents the complex conjugate of the wavelet basis function; τ is the time variable in the integration process, used to traverse all time points of the signal; is the wavelet ridge energy integral, which integrates the wavelet basis energy along the energy ridge line (the path with the most concentrated energy) in the time-frequency plane; is the attenuation factor, dynamically adjusting the energy contributions at different scales s ; ξ is the attenuation coefficient, adjusting the weights of energies at different scales s ; |Re( )| takes the absolute value of the real part, reflecting the transient energy distribution of the signal and ensuring the non-negativity of the energy.

[0058] The parameter system of wavelet ridge energy focuses on the main energy components of slip events through scale selection, band weighting and dynamic fusion, and by integrating along the ridge line, ignoring noise and irrelevant signals, improving detection sensitivity and robustness. High-sensitivity capture of slip transient characteristics is achieved.

[0059] Differences among the three criteria

[0060] Step 1.4: Propose the self-reference slip criterion S i ( t ) Propose the self-reference slip criterion( S i ( t )) is designed to solve the limitations of single-slip-criterion detection methods by fusing multi-dimensional features and improve the slip detection ability under complex working conditions. Define :

[0061] where entropy-driven regulation is used ; λ is the sensitivity false gain, controlling the regulation sensitivity of the tension distribution entropy D ( Ti ) to the weight , realizing non-linear weight distribution, λ = 0.1 - 0.3 for low-noise systems, λ = 0.5 - 1.0 for high-dynamic systems; represents the slip index value of the i th drive line at the current t moment; The advantages of the self-reference slip criterion S i ( t ) are as follows: a. Cover all types of slips Sudden slip: The tension distribution entropy quickly responds to the drastic deviation of the tension value through the fourth moment. Asymptotic slip: The fractional derivative utilizes the non-local memory effect to accumulate small changes and avoid missed detections. Micro slip: The wavelet ridge energy locks in high-frequency transient components (such as friction vibrations) to early warn of potential risks.

[0062] b. Adapt to dynamic environments Wire-driven manipulators often face uncertainties such as sudden changes in load and variations in friction coefficients. The dynamic indicators are adaptively adjusted through weighted fusion (such as entropy-driven weights): reducing the weights of criteria sensitive to high frequencies in high-noise scenarios and enhancing the contribution of criteria with historical cumulative effects in slowly varying scenarios.

[0063] c. Suppressing false alarms and missed detections There are various interferences in the industrial environment (motor vibration, electromagnetic noise), and different criteria have significantly different sensitivities to noise. For example, the tension distribution entropy is insensitive to low-amplitude random noise, while the wavelet ridge energy can filter out interferences in non-target frequency bands. Joint decision-making with multiple criteria can reduce the false detection rate through logical fusion (such as "AND / OR" rules).

[0064] d. Integrating multi-dimensional information Combining time-domain (fractional order), frequency-domain (wavelet), and statistical-domain (fourth moment) features: comprehensively considering the impact of slip on the overall state of tension, the time evolution information of the dynamic process, the moment and frequency band when slip occurs, and assisting in classifying fault types (such as abnormal friction or structural failure).

[0065] Step 2: Slip alarm with dynamic threshold S i (t) is used as the slip criterion. However, to determine the occurrence of slip, a threshold judgment needs to be defined so that the manipulator can have feedback for stable grasping. To adapt to the non-stationary noise environment, a dynamic threshold based on online kernel density estimation is introduced as the threshold for slip judgment. The online kernel density estimation function is as follows:

[0066] In the formula, is the Epanechnikov kernel function, a non-negative function used to smooth data points, spreading the influence of each data point to its neighborhood and determining the shape of the density estimate; h is the bandwidth, the width parameter of the kernel function, controlling the degree of smoothing, calibrated through experiments; N is the window length of the online kernel density, the number of historical data points for kernel density estimation, which can be set experimentally by oneself; S ( t ) is the current slip index value; S (k) The set of historical slip index values { S (1), S (2),..., S ( N )}, constructing a background probability distribution model through kernel density estimation; k represents the discrete-time index, k = 1, 2, 3,..., N , representing the k-th sampling point of the historical data. Through experiments and practical applications, continuously adjusth, N , which can ensure the stability and accuracy of the criterion under different working conditions.

[0067] Set the threshold (99th percentile) is the 99th percentile of the probability density function f ( S ), that is P ( S ≤ S th ) = 0.99, in order to reduce the possibility of misjudgment and improve the robustness of the system.

[0068] Step 3: Compare the slip criterion with the dynamic threshold to judge the manipulator's action: When S ( t ) ≤ , the manipulator maintains the current grasping state; When 2 > S ( t ) > , the manipulator increases the stiffness and enhances the fit to continue grasping; When S ( t ) ≥ 2 , the manipulator brakes, stops working and alarms.

[0069] Define multi-level slip alarm: Level 1 (slip warning): S(t) > and lasts for △t 1 Level 2 (emergency braking): S(t) ≥ 2

[0070] where, △t 1 is obtained by experiment, and △t 1 = 0.8s is more ideal.

[0071] In one embodiment, a soft manipulator with adjustable stiffness is mounted on an underwater vehicle for grasping underwater organisms (such as starfish, sea urchins, jellyfish, etc.): The fingers of the manipulator in this embodiment are made of a 3D-printed TPU skeleton, with a 30 mm difference between the peaks and valleys of the serrated structure. The drive line is a ultra-high molecular weight polyethylene cable coated with graphite lubricant; the drive motor uses a DC motor to drive a wire reel, and a PTFE fixed pulley guiding system with a friction coefficient ≤ 0.1. Four micro tension sensors are embedded in the base of the finger skeleton.

[0072] For grasping underwater organisms, the soft manipulator adopts an antagonistic or composite mode.

[0073] Approaching the target: Driven by the robotic arm, the robotic hand gently approaches the target organism. During the approach, the flexibility of the robotic hand enables it to adapt to the dynamic behavior of the target organism.

[0074] Shape adjustment: Through the control of the drive mechanism, each finger of the robotic hand finely tunes the tension of the inner and outer drive lines, achieving a moderate bending of the fingers to fit the target shape. This precise bending adjustment ensures that the robotic hand can fit the organism to the greatest extent and prevent it from escaping.

[0075] Grasping stage: When the robotic hand touches the target organism, adjust the stiffness of the robotic hand from soft to moderately rigid (by tightening the drive lines). This increased rigidity ensures a firm grasp of the target organism and prevents it from detaching from the robotic hand due to water flow or its own movement.

[0076] Intelligent feedback: The miniature tension sensors embedded in the robotic hand detect the tension changes in real time and transmit them to the controller. According to the tension signal, the slip criterion calculates and the actuator dynamically adjusts the stiffness of the robotic hand to ensure a firm grasp while avoiding excessive squeezing.

[0077] In one embodiment, a soft robotic hand with adjustable stiffness is mounted on an underwater vehicle for flexible cleaning and grasping of underwater cultural relics: The fingers of the robotic hand in this embodiment are made of a 3D-printed medical-grade silicone skeleton, with a 30-mm difference between the peaks and valleys of the serrated structure. The drive lines are ultra-high molecular weight polyethylene cables coated with graphite lubricant; the drive motor uses a DC motor to drive a wire reel and a PTFE fixed pulley guiding system with a friction coefficient ≤0.1. Four miniature tension sensors are embedded in the base of the finger skeleton.

[0078] For the grasping of fragile and valuable objects in such an underwater complex environment, the soft robotic hand adopts a flexible cooperative, antagonistic or composite mode.

[0079] Approaching the target: The drive mechanism keeps the drive lines of the robotic hand in a relaxed state, making the robotic hand in a highly compliant state to approach the cultural relic and adapt to its surrounding environment.

[0080] Contact and cleaning: The robotic hand gently touches the surface of the cultural relic, and the contact pressure is monitored in real time through the tension sensor, and the stiffness is dynamically adjusted to achieve moderate cleaning. A multi-point contact method is adopted to ensure uniform distribution of the cleaning water flow and avoid local over-washing or omission.

[0081] Stiffness adjustment and adsorption control: During the cleaning process, the operator gradually increases the stiffness of the robotic hand according to the reaction of the cultural relic surface to ensure a firm adsorption and prevent the cultural relic from being displaced due to water flow fluctuations. At the same time, maintain a certain degree of flexibility to avoid scratches on the cultural relic surface caused by rigid contact.

[0082] Real-time feedback and adjustment: Through the data analysis of the tension sensor, the cleaning effect and the status of the cultural relics are evaluated in real time. If abnormal tension changes are detected (such as the movement of cultural relics or contact with irregular obstacles), the controller will immediately adjust the shape and stiffness of the manipulator to ensure safe operation.

[0083] After cleaning, the robot is adjusted to a moderate rigidity, gently wraps the artifacts, and slowly transfers them to the collection cabin or protective container. Ensure that there is space between the artifacts and the inner wall of the container during the transfer process to avoid potential damage during transportation.

[0084] In one embodiment, a soft manipulator with adjustable stiffness is used to grab large cartons in a logistics warehouse: Application objects of this embodiment: In a logistics warehouse, large cartons need to be transported. The dimensions of the cartons are 30 cm in length, 30 cm in width, and 30 cm in height, and the weight is relatively light, about 3 kg.

[0085] The fingers of the manipulator in this embodiment use 3D printed ABS skeletons, the difference between the crest and trough of the sawtooth structure is 30mm, the drive line is ultra-high molecular weight polyethylene cable, coated with graphite lubricant; the drive motor uses a DC motor to drive the winding reel, PTFE fixed pulley guide system, and the friction coefficient is ≤0.1. Four micro tension sensors are embedded in the base of the finger skeleton.

[0086] For the grasping of such large objects, the soft manipulator adopts a compound motion mode. Since the carton is large, the manipulator needs a larger reach. The specific steps are as follows: Step 1: The driving motor 31 controls the inner and outer driving wires 12 and 13 to relax, so that the finger 1 is in a natural extension state to cover a larger area of ​​the carton.

[0087] Step 2: By asynchronously tightening the driving wires 12 and 13, the finger 1 contracts and bends inwards, thereby achieving an embracing grasping of the carton.

[0088] In this process, the retractable and bendable characteristics of the soft manipulator come into play, as large objects require a large contact area and appropriate bending to adapt to their shape. At the same time, due to the light weight of the carton, the low stiffness of the soft manipulator in the extended state is enough to support the carton, and it can be deformed appropriately according to the irregular shape of the carton to ensure stable grasping.

[0089] In one embodiment, a soft manipulator with adjustable stiffness is used for chip grabbing in electronic component assembly: The present embodiment is applicable to electronic component assembly workshops, which need to grasp tiny electronic chips. The chip size is about 5mm×5mm×1mm and the weight is very light, about 0.1g.

[0090] The fingers of the robotic arm in this embodiment adopt a 3D-printed medical-grade silicone skeleton, with a 10-mm difference between the peaks and valleys of the serrated structure. The drive wires are ultra-high molecular weight polyethylene cables coated with graphite lubricant. The drive motor uses a micro stepping motor to drive a wire reel and a PTFE pulley system with a friction coefficient ≤ 0.1. Four micro tension sensors are embedded in the base of the finger skeleton.

[0091] For grasping small objects, the soft robotic arm adopts an antagonistic motion mode. Since the electronic chip is very small, precise operation is required. The specific steps are as follows: Step 1: The drive motor 31 controls the coordinated movement of the inner and outer drive wires 12 and 13 to make the finger 1 in a slightly contracted state to increase the micro stiffness.

[0092] Step 2: By adjusting the drive motor 31, one of the drive wires 12 and 13 is tightened and the other is relaxed, so that the finger 1 can bend precisely inward or outward to adjust the distance and shape of the finger to fit the size of the chip.

[0093] Step 3: By adjusting the drive motor 31, the inner drive wire 12 is tightened and the finger bends inward to fit the edge of the chip.

[0094] Since the chip is very small and light, the robotic arm does not require high stiffness. In the natural elongation state or with lower stiffness, it can ensure that the chip will not be damaged during the grasping process. The tip of the finger 1 can be finely adjusted and bent according to the shape of the chip. By utilizing the flexibility and bendability of the skeleton 11, gentle grasping of small objects can be achieved.

[0095] In one embodiment, a soft robotic arm with adjustable stiffness is used for handling heavy objects at a construction site: The application object of this embodiment: At a construction site, it is necessary to handle precast concrete blocks with a size of 30 cm in length, 20 cm in width, 15 cm in height, and a weight of about 5 kg.

[0096] The fingers of the robotic arm in this embodiment adopt 3D-printed PLA, with a 30-mm difference between the peaks and valleys of the serrated structure. The drive wires are ultra-high molecular weight polyethylene cables coated with graphite lubricant. The drive motor uses a high-torque servo motor to drive a wire reel and a PTFE pulley system with a friction coefficient ≤ 0.1. Four industrial-grade tension sensors are embedded in the base of the finger skeleton.

[0097] For grasping the precast block, the soft robotic arm adopts a coordinated motion mode. The specific steps are as follows: Step 1: The drive motor 31 controls the inner and outer drive wires 12 and 13 to relax, so that the finger 1 is in a natural elongation state to cover a larger area of the precast block.

[0098] Step 2: Use the drive motor 31 to simultaneously tighten the drive wires 12 and 13, causing the finger 1 to contract. At this time, the sawtooth structure 112 reaches a minimum of approximately 6 N·mm, and the stiffness of the finger 1 becomes higher.

[0099] The high stiffness can provide sufficient supporting force to grasp an object weighing 5 kg. During the grasping process, since the shape of the precast block is relatively regular, the finger 1 can remain in a contracted state and use the higher stiffness to bear the weight of the precast block.

[0100] Step 3: The soft manipulator can further adjust the drive wires 12 and 13 according to the shape of the precast block to ensure full contact between the finger 1 and the surface of the precast block, achieve stable grasping, and prevent the precast block from slipping during transportation.

[0101] Step 4: Make a judgment on the corresponding execution action based on the comparison between the slip criterion and the set threshold.

[0102] Step 5: Relax the drive wires in stages and place the precast block smoothly.

[0103] In one embodiment, an adjustable-stiffness soft manipulator is used to grasp a thin food packaging bag: The application object of this embodiment: In a food packaging workshop, it is necessary to grasp a thin food packaging bag with a length of 15 cm, a width of 10 cm, and a weight of about 5 g.

[0104] The finger of the manipulator in this embodiment is made of 3D-printed ABS, with a difference of 20 mm between the peaks and valleys of the sawtooth structure. The drive wires are ultra-high molecular weight polyethylene cables coated with graphite lubricant; the drive motor uses a micro stepping motor to drive a wire reel, and a PTFE pulley system with a friction coefficient ≤ 0.1. The base of the finger skeleton is embedded with a micro tension sensor.

[0105] For grasping such light objects, the soft manipulator adopts an antagonistic motion mode. The specific steps are as follows: Step 1: Fine-tune the tightening and relaxation states of the drive wires 12 and 13 through the drive motor 31 so that the finger 1 can gently bend inward.

[0106] Since the food packaging bag is very light, the manipulator does not require high stiffness. In the state of natural elongation or slight contraction, the lower stiffness can avoid damaging the packaging bag during the grasping process.

[0107] Step 2: Adjust the bending angle and spacing of the finger 1 according to the size of the packaging bag for precise adjustment. Utilize the compliance of the soft manipulator to ensure stable grasping of the edge of the packaging bag without cutting the packaging bag or spilling the food inside the packaging bag.

[0108] In one embodiment, a soft robotic manipulator with adjustable stiffness is used for a drone to hang on a tree branch and stay: The application object of this embodiment: The robotic manipulator is used for a drone to assist the drone in hanging on tree branches of different thicknesses to verify the application scenario of staying.

[0109] In this embodiment, the fingers of the robotic manipulator are made of 3D-printed ABS material, the difference between the peaks and valleys of the serrated structure is 30 mm, the drive wires are ultra-high molecular weight polyethylene cables coated with graphite lubricant; the drive motor uses a high-torque servo motor to drive the wire reel, and the PTFE pulley system has a friction coefficient ≤ 0.1. Four micro tension sensors are embedded in the base of the finger skeleton.

[0110] (1) Adaptation to tree branches of different thicknesses Thinner tree branches (diameter about 2 - 5 cm). When the drone approaches a thinner tree branch, for thinner tree branches, a relatively gentle and precise grasping action is required. The soft robotic manipulator adopts a composite motion and antagonistic motion mode. The specific steps are as follows: Step 1: Asynchronously control the tension state of the drive wire 12 and the drive wire 13 through the drive motor 31. The tensioning stroke of the inner drive wire 12 is greater than that of the outer drive wire 13; make the finger 1 start to bend inward and approach the tree branch.

[0111] Due to the flexibility of the material of the finger skeleton 11 of the finger 1, the finger 1 can bend within the range of ±180°, and can fit well with the circular contour of the thinner tree branch.

[0112] Step 2: According to the specific position and shape of the tree branch, continue to finely adjust the tension of the drive wire 12 and the drive wire 13 to make the finger 1 tightly hold the tree branch.

[0113] In this process, due to the thinner tree branch, a very high stiffness is not required. When the finger 1 is in the state of natural elongation or slightly contracted, a lower stiffness can not only ensure stable contact with the tree branch, but also will not damage the tree branch due to too high stiffness.

[0114] Medium-thickness tree branches (diameter about 5 - 10 cm). For medium-thickness tree branches, after the drone approaches the tree branch, the soft robotic manipulator adopts a composite motion mode. The specific steps are as follows: Step 1: Control the drive wire 12 and the drive wire 13 to be in a relaxed state through the drive motor 31 to increase the contact range with the tree branch.

[0115] Step 2: Control the asynchronous tensioning of the drive wire 12 and the drive wire 13, so that the drive wire 12 is tensioned more than the drive wire 13, and make the finger 1 contract and bend inward at the same time.

[0116] Since the serrated structure 112 of finger 1 can change its stiffness when contracting, as finger 1 contracts, the stiffness gradually increases. This moderately increased stiffness can better adapt to the relatively greater support requirements of medium-thick tree branches, ensuring stable hanging on the tree branches. At the same time, the bendable characteristic of finger 1 can still make it fit well with the shape of the tree branch, ensuring sufficient contact with the tree branch and preventing the drone from slipping due to shaking.

[0117] Relatively thick tree branches (with a diameter of about 10 - 25 cm). When facing relatively thick tree branches, the soft manipulator uses a coordinated motion mode. The specific steps are as follows: Step 1: Control the synchronous tensioning of drive wire 12 and drive wire 13 through drive motor 31 to make the fingers contract and increase stiffness. At this time, the serrated knot 112 is close to the shortest, and the stiffness of finger 1 becomes higher. Sufficient supporting force is provided through the high stiffness to hold the relatively thick tree branch.

[0118] Since the shape of the relatively thick tree branch is relatively regular, finger 1 can maintain a contracted state and use the relatively high stiffness to bear the weight of the drone.

[0119] Step 2: Control the inner drive wire 12 to continue to be tensioned, making drive wire 12 more tensioned than drive wire 13, so that finger 1 contracts and bends inward while contracting.

[0120] Through the fixed pulleys inside the rope guiding mechanism 2, the tensions of drive wire 12 and drive wire 13 are effectively transmitted to finger 1, ensuring that each finger 1 can evenly apply pressure to the tree branch, thus stably hanging on the relatively thick tree branch and enabling the drone to stay safely.

[0121] (2) Coping with the surface condition of the tree branch Smooth tree branch surface. Regardless of the thickness of the smooth tree branch, when finger 1 of the soft manipulator contacts the tree branch, due to its own compliance and bendability, it can adaptively adjust according to the shape of the smooth tree branch surface. During the grasping process, the friction force between the surface of finger 1 and the tree branch surface is relatively small. However, by adjusting the bending degree and holding force of finger 1, the friction force and the wrapping force of finger 1 on the tree branch can be used to achieve stable hanging. For example, when using the antagonistic motion mode on a relatively thin smooth tree branch, the bending degree of finger 1 can be appropriately increased to increase the contact area between finger 1 and the tree branch, thereby increasing the friction force and the wrapping force to ensure the drone stays on the smooth tree branch.

[0122] Rough branch surface. For a rough branch surface, when the soft manipulator grasps it, the rough surface can provide greater friction. In this case, the soft manipulator can select an appropriate motion mode according to the thickness of the branch. For example, on a medium-thick rough branch, when using the composite motion mode, during the contraction and bending of finger 1, it can better embed into the rough texture of the branch surface. At the same time, due to the certain flexibility of the material of finger 1, it can adapt to the irregular shape of the rough surface, avoiding unstable grasping caused by protrusions or depressions on the branch surface. Even when the drone is affected by external disturbances (such as being blown by the wind), the friction and embedding effect between the rough branch surface and finger 1 can effectively prevent the drone from slipping.

[0123] (3) Stability when the drone stays Adjust the stiffness to maintain stability. During the process when the drone is hanging on the branch and staying, it may be affected by the blowing of the wind or its own slight shaking. The soft manipulator can maintain stability by adjusting the stiffness of finger 1. If the drone starts to shake, the soft manipulator can further tighten or loosen the driving wires 12 and 13 through the driving motor 31 to change the stiffness of finger 1. For example, when the drone is affected by lateral wind force, increasing the stiffness of finger 1 can improve the grasping force on the branch and prevent the drone from being blown off the branch. On the contrary, if the shaking is small and mainly vertical vibration, appropriately reducing the stiffness of finger 1 can enable the manipulator to better buffer the vibration and keep the drone staying stably.

[0124] Adapt to the shaking of the branch and keep in sync. The branch itself may also shake in the wind or have slight swings due to other external forces. The bendable and stretchable characteristics of the soft manipulator enable it to adapt to this kind of shaking of the branch. Since finger 1 can bend and stretch within a large range, when the branch shakes, finger 1 can adjust its own shape and posture along with the swing of the branch and always maintain close contact with the branch. For example, when the branch swings to the left, finger 1 can bend and stretch by appropriate adjustment of the driving wires 12 and 13 to follow the movement of the branch, avoiding the drone from slipping due to the incoordination between the branch and the manipulator, thus ensuring the safe and stable stay of the drone on the branch.

[0125] In the embodiment, the basis for maintaining the stability of the grasp is the proposed slip criterion and slip detection, that is, whether there is relative sliding between the mechanical claw and the branch, and then relevant actions are generated.

[0126] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention without departing from the principles and purposes of the present invention.

Claims

1. A soft manipulator with adjustable stiffness, characterized in that: It includes an action execution component composed of a plurality of fingers arranged along the circumferential direction, and the action execution component is controlled by a driving mechanism to realize multi-modal motion; The driving mechanism comprises a driving motor and a winding reel mounted on its output shaft; The finger comprises a sawtooth frame and at least two driving wires running through the frame along the length direction; the frame is formed in one piece with a flexible material, and holes are provided inside and outside the frame for the driving wires to pass through; one end of the driving wire is fixed to a winding reel, and the other end passes through the hole in the finger frame and is connected to the end of the finger, and a miniature tension sensor is installed at the end of the finger for real-time detection of the tension value of the driving wire; By controlling the tension or relaxation of the driving wire through the driving mechanism, the length or curvature of the finger skeleton changes, thereby achieving the extension and retraction of the finger, dynamic adjustment of stiffness and adjustment of the bending angle.

2. A soft manipulator with adjustable stiffness according to claim 1, characterized in that: The driving mechanism and the action execution component are connected by a rope guiding mechanism. The rope guiding mechanism serves as a base for fixing the roots of each finger and is provided with a plurality of fixed pulleys and guide channels for guiding the path of the driving line and reducing friction. The driving line passes through the finger channel and the guide channel at one end of the rope guiding mechanism in sequence and enters the housing of the rope guiding mechanism. After passing through a plurality of fixed pulleys, it passes through the guide channel at the other end of the rope guiding mechanism to reach the driving mechanism and is connected to the winding reel.

3. A soft manipulator with adjustable stiffness according to claim 1, characterized in that: The finger includes a sawtooth structure body and bases installed at both ends of the body. The sawtooth structure is composed of N continuous "V"-shaped structures, wherein the angle between the two side walls of the "V" shape is 90°; the base at one end serves as the root of the finger and is installed on the rope guide mechanism, and the base at the other end serves as the end of the finger, in which micro-tension sensors are arranged in the same number as the driving lines and in a one-to-one correspondence.

4. The soft manipulator with adjustable stiffness according to claim 1, characterized in that: The action execution component includes three fingers evenly distributed along the circumference, and two channels are opened in parallel at the inner / outer sawtooth groove ends of each finger, that is, the channels are arranged in a rectangular array along the length direction in space.

5. A method for controlling a soft manipulator with adjustable stiffness according to any one of claims 1 to 4, characterized in that The specific steps are as follows: The driving mechanism controls the tension or relaxation of the inner driving wire and the outer driving wire to drive the finger to achieve multi-modal motion, which includes: Synergistic movement: Synchronously tighten or relax the inner and outer drive lines to extend and retract the fingers; Antagonistic movement: tighten one side of the drive line and relax the other side of the drive line to bend the fingers inward or outward; Compound movement: asynchronous tensioning of the medial and lateral drive lines to cause the fingers to contract and flex simultaneously; Real-time detection of the tension value of each drive line, and calculation of the slip criterion based on the tension signal; According to the comparison result of the slip criterion and the dynamic threshold, the stiffness of the manipulator is adjusted or a braking operation is triggered.

6. The method for controlling the gripping of a soft manipulator with adjustable stiffness according to claim 5, characterized in that: The slip criterion By fusing multi-dimensional features, the calculation formula is as follows: in, represents the weight, , λ represents the sensitivity pseudo gain; D ( Ti ) represents the tension distribution entropy, represents the fractional derivative, represents the wavelet ridge energy, is the wavelet ridge energy integral; i Indicates the drive line number; t is the current sampling time; α represents the order of the fractional derivative, α ∈(0,1); s is the scale parameter, s∈[ ], fs represents the sampling frequency, f min and f max indicates the lowest and highest attention frequencies; t' represents the translation parameter; The tension distribution entropy: quantifies the distortion of the tension distribution based on the fourth-order central moment and variance difference of the tension value in the sliding window; The fractional derivative captures the slow relaxation or stick-slip oscillation of the tension signal through non-integer derivatives; The wavelet ridge energy is used to extract transient high-frequency energy in the tension signal and locate slip events.

7. The method for controlling the gripping of a soft manipulator with adjustable stiffness according to claim 6, characterized in that: The dynamic threshold It is determined by the 99% quantile of the online kernel density estimate, calculated as follows: Among them, Q 99 Table 99% quantile; online kernel density f (S) The estimated function expression is as follows: in, represents the Yepanechnikov kernel function; h Indicates bandwidth; N represents the window length of the online kernel density; S ( t ) represents the current slip index value; S (k) Represents the historical slip index value set { S (1), S (2), ... , S ( N )}, k represents a discrete time index, k =1,2,3,…, N .

8. The method for controlling the gripping of a soft manipulator with adjustable stiffness according to claim 7, characterized in that: The comparison between the slip criterion and the dynamic threshold is as follows: when S ( t )≤ , the robot still maintains the current grasping state; When 2 > S ( t )> , the manipulator improves rigidity, enhances fit and continues to grasp; when S ( t )≥2 , the robot brakes, stops working and sounds an alarm.

9. The method for controlling the gripping of a soft manipulator with adjustable stiffness according to claim 5, characterized in that: The stiffness of the finger is dynamically adjusted by the tension of the drive line: in coordinated motion, the stiffness increases when the finger contracts and decreases when it extends; in compound motion, the stiffness and bending angle are adjusted simultaneously by asynchronously tightening the drive line.

10. The method for controlling the gripping of a soft manipulator with adjustable stiffness according to claim 5, characterized in that: The tension value of the driving line is detected in real time by a micro tension sensor embedded in the finger base and transmitted to the controller for recursive calculation, and the tension mean and variance are updated in real time to avoid repeated calculations and improve the response speed.

Images