A wrist joint dual-axis motion detection flexible sensor and method and a robotic arm

Abstract

The present invention discloses a flexible sensor and method for detecting dual-axis motion of a wrist joint and a mechanical arm; the sensor comprises a light-proof coat, a protective layer, a strip of transparent conductive hydrogel, a light intensity sensor, a light source and a resistance detection module; the transparent conductive hydrogel is arranged in the inner cavity of the light-proof coat; a protective layer is arranged between the inner wall of the light-proof coat and the transparent conductive hydrogel; the light intensity sensor and the light source are respectively arranged at both ends of the strip of transparent conductive hydrogel. The resistance detection module is connected to both ends of the strip of transparent conductive hydrogel. The longitudinal section of the strip of transparent conductive hydrogel is a rectangle with unequal length and width. The present invention uses the light transmittance and resistance changes of the strip of transparent conductive hydrogel with inconsistent cross-sectional length and width when it is bent and stretched, and the structure obtains the rotation direction and rotation angle motion detection of the wrist joint, thereby simplifying the structural complexity of the sensor for dual-degree-of-freedom motion detection of the wrist joint.

Description

Technical Field

[0001] The present invention belongs to the field of motion detection of flexible sensors and the technical field of mechanical arms, and in particular relates to a flexible sensor and method for dual-axis motion detection of a wrist joint and a mechanical arm. Background Art

[0002] Flexible sensors have a wide range of applications in the field of motion detection. Their main advantage is that they can adapt to various complex shapes and dynamic environments. Due to their flexibility, these sensors are widely integrated into detectable devices to monitor human motion, posture changes and motion status in real time, thereby providing accurate data support. At present, there are problems such as complex structure, large size and difficulty in signal decoupling in the structural design of flexible sensors. At present, the structural design of flexible sensors mainly lies in the superposition of various single sensor mechanisms, which makes the overall structure complex and large in size, and has certain limitations on its application. Among them, when it is necessary to sense multiple different physical quantities, flexible sensors will have problems such as signal crosstalk and transmission processing difficulties, which will affect the detection accuracy.

[0003] Flexible sensors are also of great significance in the process of detecting the motion of the robotic arm. They can accurately detect the accurate motion of the robotic arm and provide data support for the control of the robotic arm. Furthermore, in the design of the robotic arm, there are still many problems in the design of a bionic robotic arm that can completely imitate the motion of the human forearm and wrist joint. The traditional serial robotic arm joints are equipped with servos or motors to drive the wrist motion. The spatial structure of the wrist joints is large and bulky, and it cannot replicate the motion of the human wrist well. Parallel robotic arms generally use spherical parallel mechanisms, which have complex structures, small working spaces, and are prone to interference.

[0004] At present, most of the designs of robotic arms use cylinders for driving. The design of the mechanism is complex, prone to leakage, noisy and difficult to control accurately. Some of them replace the pneumatic drive with a hydraulic drive, which makes the drive have the characteristics of high power density and smooth movement, but requires a liquid medium, the system is complex and prone to leakage. In addition, the intelligent material drive has a simple structure and low energy consumption, but the movement speed is slow and the current material performance is limited. Finally, some wrist joints driven by electric motors, in which the placement of the reducer occupies a large volume, make the overall structure large and do not meet the requirements of simulation. At the same time, in the five-finger drive of the human hand, the arrangement of the drive is also a major problem. Placing the finger drive inside the arm, reducing the volume of the palm and reducing the wrist load is favored by designers. However, this design causes the movement of the fingers and the movement of the wrist to interfere with each other, which will affect the drive and control of the fingers when the wrist moves. Summary of the invention

[0005] The object of the present invention is to provide a wrist joint biaxial motion detection flexible sensor and method and a robotic arm.

[0006] In the first aspect, the present invention provides a flexible sensor for detecting biaxial motion of a wrist joint, which comprises a light-proof outer jacket, a protective layer, a strip of transparent conductive hydrogel, a light intensity sensor, a light source and a resistance detection module; the transparent conductive hydrogel is arranged in the inner cavity of the light-proof outer jacket; a protective layer is arranged between the inner side wall of the light-proof outer jacket and the transparent conductive hydrogel; the light intensity sensor and the light source are respectively arranged at both ends of the strip of transparent conductive hydrogel. The resistance detection module is connected to both ends of the strip of transparent conductive hydrogel. The longitudinal section of the strip of transparent conductive hydrogel is a rectangle with unequal length and width.

[0007] During operation, the two ends of the light-proof jacket are respectively fixed on the back of the hand and arm corresponding to the mechanical wrist joint under test; the wrist joint rotates in different directions and at different speeds, causing the strip of transparent conductive hydrogel to bend and stretch, causing the light intensity measured by the light intensity sensor and the resistance value measured by the resistance detection module to change. According to the changes in light intensity and resistance value, the rotation direction and angle of the wrist joint under test are obtained.

[0008] Preferably, the protective layer is made of VHB film; the VHB film wrapped outside the hydrogel can prevent the hydrogel from losing water and deteriorating, thereby ensuring the stability of the properties of the hydrogel.

[0009] Preferably, the aspect ratio of the longitudinal section of the strip-shaped transparent conductive hydrogel is (1.5-2.5):1.

[0010] In a second aspect, the present invention provides a wrist joint biaxial motion detection method, which uses the aforementioned wrist joint biaxial motion detection flexible sensor. The wrist joint biaxial motion detection method is as follows: when the measured wrist joint rotates, the wrist joint biaxial motion detection flexible sensor collects time series data of light intensity and resistance, and inputs the time series data into a wrist joint rotation detection network. The wrist joint rotation detection deep learning algorithm outputs the rotation direction and rotation angle of the wrist joint. The wrist joint rotation detection network adopts a bidirectional long short-term memory deep learning algorithm.

[0011] Preferably, the wrist joint rotation detection network includes a forward LSTM network and a backward LSTM network. Both the forward LSTM network and the backward LSTM network include multiple hidden layers. Two adjacent hidden layers are connected by residuals.

[0012] In a third aspect, the present invention provides a bionic mechanical arm, comprising an arm support, a palm support, a wrist joint connection assembly, a wrist drive mechanism, a finger drive mechanism and a plurality of bionic fingers. The inner end of the palm support is connected to the end of the arm support through the wrist joint connection assembly. The wrist drive mechanism is used to drive the palm support to perform a two-degree-of-freedom flip. Each bionic finger is connected to the palm support and is driven by the finger drive mechanism to bend and straighten.

[0013] The wrist joint connection assembly includes a wrist cross universal joint, a palm base and an arm end seat. The wrist cross universal joint is provided with a first connecting rod and a second connecting rod perpendicular to each other. The palm base is fixed on the palm support. The arm end seat is fixed on the arm support. The palm base and the first connecting rod respectively constitute a first rotation pair. The arm end seat and the second connecting rod constitute a second rotation pair.

[0014] The wrist drive mechanism includes two unilateral drive components. The unilateral drive components include a motor mounting bracket, a direct-acting motor and a ball joint housing. The motor mounting bracket and the arm bracket form a first spherical pair. The direct-acting motor is mounted on the motor mounting bracket. The ejection rod of the direct-acting motor and the side of the palm base form a second spherical pair. The rotational sphere centers of the second spherical pairs corresponding to the two unilateral drive components are respectively located on the same side of the rotation axis of the first rotation pair; the rotational sphere centers of the second spherical pairs corresponding to the two unilateral drive components are respectively located on the opposite sides of the rotation axis of the second rotation pair.

[0015] Preferably, the unilateral drive assembly further comprises a ball joint seat and a ball joint rod. The ball joint seat is fixed on the arm bracket. One end of the motor mounting bracket is fixed to one end of the ball joint rod. The connecting ball at the other end of the ball joint rod and the ball joint seat form a first spherical pair. The line connecting the rotating ball centers of the first spherical pair and the second spherical pair is parallel to or coincides with the movement direction of the push rod of the direct-acting motor.

[0016] Preferably, a flexible sensor is further included. The flexible sensor is connected between the palm support and the arm support. The flexible sensor includes a light-proof coat, a protective layer, a strip of transparent conductive hydrogel, a light intensity sensor and a light source. The transparent conductive hydrogel is arranged in the inner cavity of the light-proof coat. A protective layer is arranged between the inner side wall of the light-proof coat and the transparent conductive hydrogel. The light intensity sensor and the light source are respectively arranged at both ends of the transparent conductive hydrogel.

[0017] Preferably, the bionic finger comprises a spring and a plurality of knuckles. Each knuckle is a support knuckle and a plurality of movable knuckles which are connected in rotation from the inside to the outside. The support knuckle is fixed on the palm support; any three consecutive knuckles correspond to a linkage rod. The two ends of the linkage rod are respectively connected in rotation with the opposite ends of two non-adjacent knuckles among the three knuckles corresponding to the linkage rod. The rotation axis between the linkage rod and the knuckle does not coincide with the rotation axis between adjacent knuckles.

[0018] Preferably, any three consecutive finger joints correspond to two rotational joints; the axis of the linkage rod intersects with a line connecting the rotation centers of the two corresponding rotational joints.

[0019] Preferably, the wrist cross universal joint comprises an integrally formed central ring and four connecting rods. The inner ends of the four connecting rods are connected to the outer circumference of the central ring. The four connecting rods are two first connecting rods facing each other and coaxial, and two second connecting rods facing each other and coaxial.

[0020] The inner end of the palm support is provided with a rope threading hole. The rope threading hole is aligned with the center hole on the center ring. A clearance hole connected to the internal space of the bionic finger is provided on the support knuckle; the finger drive mechanism includes a servo mounting platform and multiple single-finger drive components. The servo mounting platform is fixed on the arm support. The number of single-finger drive components is equal to that of bionic fingers and corresponds one to one. The single-finger drive component includes a finger-driven servo, a winch and a rope. The finger-driven servo is installed on the servo mounting platform; the winch is fixed on the output shaft of the finger-driven servo; one end of the rope is fixed on the winch. The rope passes through the center hole of the center ring, the rope threading hole on the palm support, the clearance hole on the support knuckle corresponding to the bionic finger, and the inner cavity of each movable knuckle, and the outer end of the rope is fixed to the movable knuckle at the end of the corresponding bionic finger.

[0021] Preferably, both ends of the center hole of the center ring are provided with arc chamfers.

[0022] Preferably, there are five bionic fingers; one of the bionic fingers is connected to the side of the palm, with two movable knuckles; the other four bionic fingers are connected side by side to the outer end of the palm, with three movable knuckles each.

[0023] Preferably, a shell structure is fixedly sleeved on the outer side of the arm support, and the palm base and the shell structure of the arm support are sealed and connected via an elastic structure.

[0024] The process of the aforementioned driving method of the bionic robotic arm is as follows:

[0025] When the palm support needs to be driven to flip up and down, the two direct-acting motors are controlled to extend or shorten synchronously to drive the palm support to rotate around the rotation axis of the first rotation pair.

[0026] When the palm support needs to be driven to swing left and right, one of the direct-acting motors is synchronously controlled to extend and the other direct-acting motor is controlled to shorten, thereby driving the palm support to rotate around the rotation axis of the second rotation pair.

[0027] During the movement of the palm support, the wrist joint dual-axis motion detection flexible sensor collects light intensity signals and resistance signals in real time, and determines the rotation direction and rotation angle of the palm support based on the measured light intensity signals and resistance signals.

[0028] When it is necessary to drive part or all of the bionic fingers to move, the finger driving mechanism drives the corresponding bionic fingers to bend or straighten.

[0029] The specific beneficial effects of the present invention are:

[0030] 1. The present invention uses a strip of transparent conductive hydrogel with inconsistent cross-sectional length and width to change the light transmittance and resistance when it is bent and stretched, and the structure obtains the rotation direction and rotation angle motion detection of the wrist joint, thereby simplifying the sensor structure complexity of the wrist joint double-degree-of-freedom motion detection.

[0031] 2. The present invention uses a bidirectional long short-term memory deep learning algorithm to decouple the timing signals measured by the flexible sensor for dual-axis motion detection of the wrist joint, and can quickly and accurately obtain the rotation direction and angle of the wrist joint; and the present invention uses residual connections between hidden layers; improves the training process of BiLSTM, accelerates convergence, improves performance, and reduces the risk of overfitting, so that the model can more effectively learn complex function mappings.

[0032] 3. The present invention integrates the finger drive mechanism on the arm support to control the bionic finger by wire; the present invention opens a center hole on the wrist cross universal joint, passes the rope through the center hole, and makes the rope pass through the rotation center of the palm support, thus avoiding the situation where the palm support rotation affects the bionic finger by wire control. While all power components are integrated on the arm support, the wrist joint drive and the finger joint drive are completely decoupled, the control logic of the multi-degree-of-freedom bionic robotic arm is simplified, and the control accuracy is improved. On the one hand, this design reduces the size of the palm to better fit the actual human hand size, and on the other hand, it reduces the weight of the palm part, reduces the load and drive inertia of the wrist drive, and prevents damage to the wrist drive.

[0033] 4. The present invention can complete various movements and functions of the human arm. It has the advantages of high efficiency, simple structure, small size and light weight, which facilitates the integration between the bionic mechanical arm and the robot, and helps the bionic robot to better realize the movement and grasping of the arm. At the same time, the present invention realizes precise control of wrist movement by highly integrating wrist control with external flexible sensors, and the design of the sensor can realize precise movement detection of wrist bidirectional bending. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a schematic diagram of the structure of Example 1 of the present invention.

[0035] Figure 2 This is a schematic diagram of the overall structure of Example 2 of the present invention.

[0036] Figure 3 Schematic diagram of the internal structure of the bionic robotic arm in Example 2 of the present invention.

[0037] Figure 4Schematic diagram of the combination of the wrist joint connection assembly and the wrist drive mechanism in Example 1 of the present invention.

[0038] Figure 5 This is the first cross-sectional schematic diagram of the wrist joint connection assembly in Example 1 of the present invention.

[0039] Figure 6 This is a second cross-sectional schematic diagram of the wrist joint connection assembly in Example 1 of the present invention.

[0040] Figure 7 It is a cross-sectional schematic diagram of the thumb in Example 1 of the present invention.

[0041] Figure 8 It is a cross-sectional schematic diagram of the index finger in Example 1 of the present invention.

[0042] Fig. 9 This is a side view of the thumb in Example 1 of the present invention.

[0043] Fig.10 It is a side schematic diagram of the index finger in Example 1 of the present invention.

[0044] Figure numerals: 1, arm support; 2, palm support; 3, wrist joint connection assembly; 3-1, wrist cross universal joint; 3-1-1, center ring; 3-1-2, first connecting rod; 3-1-3, second connecting rod; 3-2, palm base; 3-3, arm end seat; 4, wrist drive mechanism; 4-1, ball joint seat; 4-2, ball joint rod; 4-3, motor mounting bracket; 4-4, direct-acting motor; 4-5, ball joint housing; 5, finger drive mechanism; 5-1, servo mounting platform; 5-2, finger drive servo; 5-3, capstan; 5-4, rope; 6. Thumb; 6-1. Thumb support; 6-2. Proximal knuckle of thumb; 6-3. Distal knuckle of thumb; 7. Index finger; 7-1. Index finger support; 7-2. Proximal knuckle of index finger; 7-3. Middle knuckle of index finger; 7-4. Distal knuckle of index finger; 8. Middle finger; 9. Ring finger; 10. Little finger; 11. Spring; 12. Support knuckle; 13. Movable knuckle; 14. Linkage rod; 15. Flexible sensor; 15-1. Light-proof coat; 15-2. Protective layer; 15-3. Strip transparent conductive hydrogel; 15-4. Light intensity sensor; 15-5. Light source. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0046] Example 1

[0047] A wrist joint dual-axis motion detection flexible sensor is used to detect the ability to flip up and down and swing left and right. Flipping up means that the palm flips toward the back of the hand; flipping down means that the palm flips away from the back of the hand; swinging left means that the palm swings toward the thumb; swinging right means that the palm swings toward the little finger.

[0048] like Figure 1 As shown, the flexible sensor 15 includes a light-proof coat 15-1, a protective layer 15-2, a strip-shaped transparent conductive hydrogel 15-3 (length 45 mm, width 6 mm, height 3 mm), a light intensity sensor 15-4, a light source 15-5 and a resistance detection module. The light source 15-5 adopts a patch LED lamp; the light intensity sensor 15-4 adopts a patch photodiode. The protective layer 15-2 adopts a VHB film to prevent the strip-shaped transparent conductive hydrogel 15-3 from losing water. The rectangular transparent conductive hydrogel is arranged in the inner cavity of the light-proof coat 15-1. A protective layer 15-2 is arranged between the inner side wall of the light-proof coat 15-1 and the transparent conductive hydrogel. The light intensity sensor 15-4 and the light source 15-5 are respectively arranged at both ends of the transparent conductive hydrogel.

[0049] The light-proof jacket 15-1 is formed by casting to prevent the influence of external light sources on the optical signal. For easy installation, rectangular mounting blocks are cast at both ends of the light-proof jacket 15-1 of the flexible sensor 15. During the detection process, the two rectangular mounting blocks are fixed to the palm and the arm respectively, and the axis of the flexible sensor 15 is staggered from the rotation center of the palm, so that the turning and swinging of the palm can cause the flexible sensor 15 to expand and contract.

[0050] The longitudinal section of the strip transparent conductive hydrogel 15-3 is a rectangle with unequal length and width. In this embodiment, the longitudinal section of the strip transparent conductive hydrogel 15-3 has a wide side (i.e., length side) parallel to the left-right swinging direction of the palm, and a narrow side (i.e., width side) parallel to the up-and-down flipping direction of the palm.

[0051] When the flexible sensor 15 is bent bionic, the light resistance between the light intensity sensor 15-4 and the light source 15-5 changes (specifically, the number of light reflections changes), and the resistance of the strip transparent conductive hydrogel 15-3 also changes bionic. Therefore, the bending direction and bending angle of the palm relative to the arm can be inferred based on the light intensity measured by the light intensity sensor 15-4 and the resistance value measured by the resistance detection module.

[0052] The resistance detection module is connected to both ends of the strip transparent conductive hydrogel 15-3, and is used to detect the resistance value change of the strip transparent conductive hydrogel 15-3. The resistance detection module obtains the resistance value change of the strip transparent conductive hydrogel 15-3 by connecting a voltage divider resistor in series with the strip transparent conductive hydrogel 15-3 and detecting the resistance value of the voltage divider resistor.

[0053] When the bionic robot arm swings forward, the sensor undergoes wide-side bending and stretching motion; when the bionic robot arm swings backward, the sensor undergoes wide-side bending motion; when the bionic robot arm swings left and right, the sensor undergoes narrow-side bending motion. In uniform motion, when the sensor undergoes stretching deformation, the resistance of the hydrogel inside it increases with the increase in the amount of stretching; when the sensor undergoes wide-side bending and narrow-side bending, due to the rectangular design of the optical channel, the change range and change rate of the signal generated by the wide-side bending and narrow-side bending are different. According to the above signal characteristics of the sensor. Design the corresponding signal acquisition circuit to perform the signal changes of the photoelectric signal and the hydrogel signal. Input the collected signal into the improved BiLSTM algorithm for decoupling and prediction. Among them, when the sensor is in a stretch-bending composite deformation, the hydrogel signal is sensitive to the stretching length, but almost insensitive to bending. Although the photoelectric signal is sensitive to both stretching and bidirectional bending, the bidirectional bending angle cannot be directly obtained due to the nonlinear signal response of the photoelectric signal.

[0054] Since the height and width of the longitudinal section of the strip transparent conductive hydrogel 15-3 are not equal, when the swinging speed of the palm support 2 is fixed, the light intensity change rate measured by the light intensity sensor 15-4 and the resistance value change rate of the resistance detection module can be used to determine whether the palm support 2 is flipping up and down or swinging left and right, and the rotation angle can be inferred.

[0055] In this embodiment, the flexible sensor 15 combines the resistance change rate with the photoelectric sensing mechanism, and can collect bidirectional bending and stretching deformation. In the application of the bionic mechanical wrist, the combined algorithm can decouple and predict the bending angle of the bionic mechanical wrist.

[0056] The detection method of the wrist joint dual-axis motion detection flexible sensor is as follows:

[0057] Step 1. Construct a data set using the experimental data obtained from the test on the standard test bench. Each sample in the data set is the time series data of the light intensity and resistance value collected in real time by the signal acquisition circuit for the flexible sensor for dual-axis motion detection of the wrist joint; in this embodiment, the time series data is specifically the light intensity and resistance value measured 1s before the current moment; the sampling period of the flexible sensor for dual-axis motion detection of the wrist joint is 10Hz. Each sample has a label. The label is the real-time rotation direction and rotation angle of the wrist joint. The rotation speed of the wrist joint is kept constant during the experimental data collection; the rotation speed is the set rotation speed of the wrist joint of the manipulator.

[0058] A training set and a validation set are constructed based on the dataset. The training set is used to train the model, and the validation set is used to evaluate the accuracy of the model.

[0059] Step 2: Construct a wrist rotation detection network. The wrist rotation detection network uses a bidirectional long short-term memory deep learning algorithm (BiLSTM). The bidirectional long short-term memory deep learning algorithm can more effectively capture contextual information and understand the long-term dependencies of sequences when examining sequence data. It includes a forward LSTM network and a backward LSTM network. Both the forward LSTM network and the backward LSTM network include multiple hidden layers. Two adjacent hidden layers are connected by residuals.

[0060] Each hidden layer has its own input gate, forget gate, and output gate, which together improve the model's ability to understand and represent sequence data. When using BiLSTM, two independent LSTM networks can run simultaneously. The forward LSTM network learns the dependencies from the beginning of the sequence to the current time step, while the backward LSTM network learns the dependencies from the end of the sequence to the current time step.

[0061] Forward LSTM calculation: The hidden state of the forward LSTM network and The cell state is initialized to zero or a random value. At each time step from 1 to T The output and state of the forward LSTM network are calculated. The calculation process of the forward LSTM network is as follows:

[0062] 1) Forget Gate :

[0063] Formula (1)

[0064] in, Indicates that through the previous memory, Indicates discarding the previous memory. Indicates the current input.

[0065] 2) Input Gate and candidate cell status :

[0066] Formula (2)

[0067] Formula (3)

[0068] 3) Update unit status :

[0069] Formula (4)

[0070] 4) Output Gate and hidden state :

[0071] Formula (5)

[0072] Formula (6)

[0073] 5) Residual Connection:

[0074] Formula (7)

[0075] in, Represents the cell state of LSTM; , , and are the weight matrices of the forget gate, input gate, candidate unit state, and output gate respectively; , , , , and Represent the forget gate, input gate, output gate, candidate unit, input and recursive information respectively; , , and is the bias vector; σ is the sigmoid activation function; tanh is the activation function; the operator “·” represents the point multiplication of two vectors; represents the LSTM residual link; is the output of the LSTM layer, is the input of this layer.

[0076] The backward LSTM network calculation is similar to the forward LSTM network, but in the opposite order. The backward LSTM network calculation is calculated from the end of the sequence to the beginning.

[0077] The input gate determines whether to include the current input information in the cell state and controls the degree of influence of the input information on the cell state. The forget gate chooses which data in the cell state of the previous time step to delete and whether to ignore this data. The output gate chooses which information to send to the next layer of the network and how much of the cell state of the current time step to transfer to the hidden state. These three gates work together to identify and retain patterns and structures in sequential data, helping them better control the flow of information. Through BiLSTM, the hidden state of the forward LSTM network and the hidden state of the backward LSTM network at the corresponding moment are combined to obtain the following results.

[0078] Z t = W z [ h t ⃗ , h t ⃖ ] Formula (8)

[0079] in, represents the output of the BiLSTM unit, represents the weight vector, and Represent the hidden states output through the forward and reverse transmission paths, respectively.

[0080] This embodiment improves the BiLSTM algorithm by performing residual connections on two independent LSTM networks. Residual connections can improve the training process of the deep learning model BiLSTM, accelerate convergence, improve performance, and reduce the risk of overfitting, so that the model can learn complex function mappings more effectively.

[0081] Step 3: Use the training set to train the wrist rotation detection network. The number of hidden layers of the forward LSTM network and the backward LSTM network in the wrist rotation detection network is 2; the input feature of each hidden layer is 2, and the state feature is 200. Select "adam" as the solver, the maximum number of iterations is 100, and the initial learning rate is set to 0.001. After learning the deep learning algorithm, the corresponding deformation (stretching amount, width bending amount, and height bending amount) of the sensor during compound deformation can be output, and the movement direction and amount of the bionic mechanical arm can be reflected. Further, the detected arm movement amount can be used as feedback information to achieve precise control of wrist movement. Among them, the training error decreases with the increase of the model Epoch. For example, when Epoch=90, the training error (loss) of S1 is 0.0013. The training error (loss) of S2 is 0.0057. By comparing the prediction results of the validation set with the original data of the validation set, the prediction effect is determined according to the goodness of fit. The goodness of fit of the algorithm prediction value is as high as 99%.

[0082] Step 4: Use the wrist joint dual-axis motion detection flexible sensor to collect light intensity signals and resistance signals in real time, and input the real-time updated timing data of the light intensity signals and resistance signals into the wrist joint rotation detection network; the wrist joint rotation detection network outputs the current wrist joint rotation direction and rotation angle. The rotation speed setting value of the measured wrist joint is consistent with the rotation speed set in step 1.

[0083] Example 2

[0084] like Figure 2 and Figure 3As shown, a bionic mechanical arm includes an arm support 1, a palm support 2, a wrist joint connection component 3, a wrist drive mechanism 4, a finger drive mechanism 5, a thumb 6, an index finger 7, a middle finger 8, a ring finger 9, a little finger 10 and a flexible sensor 15. The inner end of the palm support 2 is connected to the end of the arm support 1 through the wrist joint connection component 3, so that the palm support 2 can rotate with two degrees of freedom relative to the arm support 1. The outer side of the arm support 1 is sleeved and fixed with a shell structure. The edge of the palm base 2 is connected to the edge of the end of the shell structure of the arm support 1 through a cylindrical elastic silicone, which ensures flexibility and sealing when the wrist is rotated. The flexible sensor 15 is connected between the palm support 2 and the arm support 1. In this embodiment, the two rectangular mounting blocks on the flexible sensor 15 are respectively fixed to the back of the palm support 2 and the back of the shell structure of the arm support 1.

[0085] The wrist drive mechanism 4 is used to drive the palm support 2 to perform two-degree-of-freedom flipping. The thumb 6 is connected to the side of the palm support 2; the index finger 7, the middle finger 8, the ring finger 9 and the little finger 10 are connected side by side to the outer end of the palm support 2; the distribution of the five fingers is consistent with the left or right hand of the human body. The thumb 6 has two joints. The index finger 7, the middle finger 8, the ring finger 9 and the little finger 10 have the same structure and all have three joints. The finger drive mechanism 5 is used to independently control the bending and straightening of the thumb 6, the palm support 2, the index finger 7, the middle finger 8, the ring finger 9 and the little finger 10.

[0086] like Figure 4 , Figure 5 and Figure 6 As shown, the wrist joint connection assembly 3 includes a wrist cross universal joint 3-1, a palm base 2 and an arm end base 1. The wrist cross universal joint 3-1 includes an integrally formed central ring 3-1-1 and four connecting rods. The inner ends of the four connecting rods are connected to the outer circumference of the central ring 3-1-1. The four connecting rods are evenly distributed along the center of the central ring 3-1-1 to form a cross structure. The four connecting rods are respectively two first connecting rods 3-1-2 that are opposite to each other and coaxial, and two second connecting rods 3-1-3 that are opposite to each other and coaxial.

[0087] Both ends of the center hole of the center ring 3-1-1 are provided with arc chamfers. The palm base 2 is fixed to the inner end of the palm support 2. The arm end seat 1 is fixed to the end of the arm support 1. The two coaxial connection holes on the palm base 2 and the two first connecting rods 3-1-2 respectively constitute the first rotation pair. The two coaxial connection holes on the arm end seat 1 and the two second connecting rods 3-1-3 respectively constitute the second rotation pair.

[0088] The wrist drive mechanism 4 includes two unilateral drive components arranged side by side. When the two unilateral drive components move in the same direction, they drive the palm support 2 to flip up and down, and when they move in the opposite direction, they drive the palm support 2 to swing left and right. The unilateral drive component includes a ball joint seat 4-1, a ball joint rod 4-2, a motor mounting bracket 4-3, a direct-acting motor 4-4 and a ball joint shell 4-5. The ball joint seat 4-1 is fixed on the arm support 1. One end of the motor mounting bracket 4-3 is fixed to one end of the ball joint rod 4-2. The connecting ball at the other end of the ball joint rod 4-2 and the ball joint seat 4-1 form a first spherical pair. The housing of the direct-acting motor 4-4 is fixed on the motor mounting bracket 4-3. A ball joint shell 4-5 is installed at the end of the ejection rod of the direct-acting motor 4-4. The ball joint shell 4-5 and the side of the palm base 2 form a second spherical pair. The line connecting the rotating ball centers of the first spherical pair and the second spherical pair is parallel to or coincides with the movement direction of the ejection rod of the direct-acting motor 4-4.

[0089] The rotation centers of the second spherical pairs corresponding to the two single-sided drive components are located on the same side of the rotation axis of the first rotation pair between the palm base 2 and the wrist cross universal joint 3-1; the rotation centers of the second spherical pairs corresponding to the two single-sided drive components are located on the opposite side of the rotation axis of the second rotation pair between the arm end seat 1 and the wrist cross universal joint 3-1.

[0090] like Figure 7 , Figure 8 , Fig. 9 and Fig.10 As shown, the thumb 6, index finger 7, middle finger 8, ring finger 9, and little finger 10 all include a spring 11 and a plurality of knuckles connected in rotation in sequence. Each knuckle is a support knuckle 12 and a plurality of movable knuckles 13 which are arranged in sequence from the inside to the outside. The support knuckle 12 is fixed on the palm support 2; any three consecutive knuckles correspond to a linkage rod 14. The two ends of the linkage rod 14 are rotationally connected to the opposite ends of two non-adjacent knuckles among the three knuckles corresponding to itself. Any three consecutive knuckles correspond to two rotating joints; the axis of the linkage rod 14 intersects with the line connecting the rotation centers of the two corresponding rotating joints, so that the two rotating joints form a linkage structure that rotates synchronously in the same direction.

[0091] The thumb 6 has two movable knuckles 13, and only one linkage rod 14 is provided; the index finger 7, middle finger 8, ring finger 9, and little finger 10 each have three movable knuckles 13, and each has two linkage rods 14. Each linkage rod 14 enables each bionic finger to rotate synchronously in the same direction.

[0092] For example, the three knuckles of thumb 6 are thumb 6 support, thumb 6 proximal knuckle and thumb 6 distal knuckle. The four knuckles of index finger 7 are index finger 7 support, index finger 7 proximal knuckle, index finger 7 middle knuckle and index finger 7 distal knuckle.

[0093] The palm support 2 is a frame structure with a cavity in the middle. A rope threading hole is provided in the middle of the inner end of the palm support 2. The rope threading hole is aligned with the center hole on the wrist cross universal joint 3-1.

[0094] The finger drive mechanism 5 includes a servo mounting platform 5-1 and five single-finger drive components. The servo mounting platform 5-1 is connected to the arm support 1 by bolts. The five single-finger drive components correspond to the thumb 6, the index finger 7, the middle finger 8, the ring finger 9, and the little finger 10 respectively. The single-finger drive component includes a finger-driven servo 5-2, a winch 5-3, and a rope 5-4. The finger-driven servo 5-2 is mounted on the servo mounting platform 5-1; a winch 5-3 is fixed on the output shaft of the finger-driven servo 5-2; and one end of the rope 5-4 is fixed on the winch 5-3. The rope 5-4 passes through the center hole on the wrist cross universal joint 3-1, the rope threading hole and the internal cavity on the palm support 2, the clearance hole on the support knuckle 12 of the corresponding finger, and the inner cavity of each movable knuckle 13 in sequence, and the end is fixed to the outermost movable knuckle 13. The finger-driven servo 5-2 drives the winch 5-3 to reel and loosen the rope 5-4 by forward and reverse rotation.

[0095] Since the rotation axes of the two rotational degrees of freedom (i.e., the first rotation pair and the second rotation pair) of the palm support 2 both pass through the center hole of the wrist cross universal joint 3-1; and the rope 5-4 passes through the center hole of the wrist cross universal joint 3-1, the two-degree-of-freedom rotation of the palm support 2 has no effect on the length of the rope 5-4 between the wrist cross universal joint 3-1 and the winch 5-3, and the rope 5-4 between the wrist cross universal joint 3-1 and the fingertips of each bionic finger; the situation in which the rotation of the palm support 2 affects the bionic finger wire control is avoided, and while all power components are integrated on the arm support 1, complete decoupling between the wrist joint drive and the finger joint drive is achieved, the control logic of the multi-degree-of-freedom bionic robotic arm is simplified, and the control accuracy is improved.

[0096] During operation, the five fingers are controlled to drive the servo 5-2 to rotate forward and backward, thereby achieving the straightening and bending of the bionic fingers and thus achieving the grasping movement of the hand.

[0097] This embodiment uses a cross universal joint to connect the palm support and the arm support, and uses two linear motors to move in the same direction and in the opposite direction to respectively realize the up and down flipping drive and left and right swinging drive of the palm support, thereby simplifying the drive design of the wrist part. At the same time, the cross universal joint of the wrist joint is directly driven by the linear motor to realize the rotation of the wrist. This design simplifies the control process, ensures the control accuracy, simplifies the structure, and makes the mechanism compact and space-saving.

[0098] In addition, the parts used in the present invention are mostly general parts, the overall structure is simple, the flexibility is good, and the human arm movement can be perfectly imitated. At the same time, a new type of flexible sensor is integrated externally to detect the movement of the mechanical wrist. The wrist mechanism has a complete structure and complete functions. Compared with other wrist mechanisms, it can provide large torque and large angle movement. The wiring is concentrated in a position that does not interfere with the structure. The structure is compact and easy to control.

[0099] The driving method of the bionic mechanical arm support 1 is as follows:

[0100] When the palm support 2 needs to be driven to flip up and down, the two direct-acting motors 4 - 4 are controlled to extend or shorten synchronously, so as to drive the palm support 2 to rotate around the rotation axis of the first rotation pair.

[0101] When it is necessary to drive the palm support 2 to swing left and right, one of the direct-acting motors 4-4 is synchronously controlled to extend and the other direct-acting motor 4-4 is shortened, so as to drive the palm support 2 to rotate around the rotation axis of the second rotation pair.

[0102] During the movement of the palm support 2, the wrist joint dual-axis motion detection flexible sensor collects light intensity signals and resistance signals in real time, and generates time series data to input into the wrist joint rotation detection network, which outputs the rotation direction and rotation angle of the palm support 2. The rotation speed setting value of the palm support 2 is consistent with the wrist joint rotation speed when the training set of the wrist joint rotation detection network is sampled.

[0103] When one or more bionic fingers need to be driven to grasp or release, the finger corresponding to the bionic finger that needs to move drives the servo 5-2 to rotate, driving the rope 5-4 to reel in or out, thereby driving the bionic finger to bend or straighten, thereby realizing the independent movement and grasping movement of each bionic finger.

Claims

1. A wrist joint biaxial motion detection method, characterized in that: The wrist joint biaxial motion detection flexible sensor used comprises a light-proof coat (15-1), a protective layer (15-2), a strip-shaped transparent conductive hydrogel (15-3), a light intensity sensor (15-4), a light source (15-5) and a resistance detection module; the transparent conductive hydrogel is arranged in the inner cavity of the light-proof coat (15-1); a protective layer (15-2) is arranged between the inner wall of the light-proof coat (15-1) and the transparent conductive hydrogel; the light intensity sensor (15-4) and the light source (15-5) are respectively arranged at two ends of the strip-shaped transparent conductive hydrogel (15-3); the resistance detection module is connected to two ends of the strip-shaped transparent conductive hydrogel (15-3); the longitudinal section of the strip-shaped transparent conductive hydrogel (15-3) is a rectangle with unequal length and width; During operation, the two ends of the light-proof jacket (15-1) are respectively fixed on the palm and arm corresponding to the wrist joint to be measured; the wrist joint rotates in different directions and at different speeds, causing the strip-shaped transparent conductive hydrogel (15-3) to bend and stretch, causing the light intensity measured by the light intensity sensor (15-4) and the resistance value measured by the resistance detection module to change, and the rotation direction and rotation angle of the wrist joint to be measured are obtained according to the changes in light intensity and resistance value; The wrist joint biaxial motion detection method is as follows: when the wrist joint to be measured rotates, the wrist joint biaxial motion detection flexible sensor collects time series data of light intensity and resistance, and inputs the time series data into the wrist joint rotation detection network; The wrist joint rotation detection network outputs the rotation direction and rotation angle of the wrist joint; The wrist rotation detection network adopts a bidirectional long short-term memory deep learning network.

2. A wrist joint biaxial motion detection method according to claim 1, characterized in that: The longitudinal section of the strip-shaped transparent conductive hydrogel (15-3) has an aspect ratio of (1.5-2.5):

1.

3. A wrist joint biaxial motion detection method according to claim 1, characterized in that: The wrist joint rotation detection network comprises a forward LSTM network and a backward LSTM network; the forward LSTM network and the backward LSTM network both comprise a plurality of hidden layers; and two adjacent hidden layers are connected via residuals.

4. A bionic mechanical arm, comprising an arm support (1), a palm support (2), a wrist joint connection component (3), a wrist drive mechanism (4), a finger drive mechanism (5) and a plurality of bionic fingers; the inner end of the palm support (2) is connected to the end of the arm support (1) via the wrist joint connection component (3); the wrist drive mechanism (4) is used to drive the palm support (2) to perform two-degree-of-freedom flipping; each bionic finger is connected to the palm support (2) and is driven by the finger drive mechanism (5) to bend and straighten; characterized in that: The bionic mechanical arm also includes a wrist joint biaxial motion detection flexible sensor; the wrist joint biaxial motion detection flexible sensor includes a light-proof coat (15-1), a protective layer (15-2), a strip-shaped transparent conductive hydrogel (15-3), a light intensity sensor (15-4), a light source (15-5) and a resistance detection module; the transparent conductive hydrogel is arranged in the inner cavity of the light-proof coat (15-1); a protective layer (15-2) is arranged between the inner side wall of the light-proof coat (15-1) and the transparent conductive hydrogel; the light intensity sensor (15-4) and the light source (15-5) are respectively arranged at two ends of the strip-shaped transparent conductive hydrogel (15-3); the resistance detection module is connected to two ends of the strip-shaped transparent conductive hydrogel (15-3); the longitudinal section of the strip-shaped transparent conductive hydrogel (15-3) is a rectangle with unequal length and width; During operation, the two ends of the light-proof jacket (15-1) are respectively fixed on the palm and arm corresponding to the wrist joint to be measured; the wrist joint rotates in different directions and at different speeds, causing the strip-shaped transparent conductive hydrogel (15-3) to bend and stretch, causing the light intensity measured by the light intensity sensor (15-4) and the resistance value measured by the resistance detection module to change, and the rotation direction and rotation angle of the wrist joint to be measured are obtained according to the changes in light intensity and resistance value; The flexible sensor (15) is connected between the palm support (2) and the arm support (1); the wrist joint connection assembly (3) comprises a wrist cross universal joint (3-1), a palm base (2) and an arm end seat (1); the wrist cross universal joint (3-1) is provided with a first connecting rod (3-1-2) and a second connecting rod (3-1-3) which are perpendicular to each other; the palm base (2) is fixed on the palm support (2); the arm end seat (1) is fixed on the arm support (1); the palm base (2) and the first connecting rod (3-1-2) respectively constitute a first rotation pair; the arm end seat (1) and the second connecting rod (3-1-3) constitute a second rotation pair; The wrist drive mechanism (4) comprises two single-sided drive components; the single-sided drive components comprise a motor mounting bracket (4-3), a direct-acting motor (4-4) and a ball joint housing (4-5); the motor mounting bracket (4-3) and the arm bracket (1) form a first spherical pair; the direct-acting motor (4-4) is mounted on the motor mounting bracket (4-3); the ejection rod of the direct-acting motor (4-4) and the side of the palm base (2) form a second spherical pair; the rotational sphere centers of the second spherical pairs corresponding to the two single-sided drive components are respectively located on the same side of the rotational axis of the first rotational pair; the rotational sphere centers of the second spherical pairs corresponding to the two single-sided drive components are respectively located on the opposite sides of the rotational axis of the second rotational pair.

5. A bionic robotic arm according to claim 4, characterized in that: The bionic finger comprises a spring (11) and a plurality of finger joints; each finger joint comprises a support finger joint (12) and a plurality of movable finger joints (13) which are connected in rotation from the inside to the outside in sequence; the support finger joint (12) is fixed on a palm support (2); any three consecutive finger joints correspond to a linkage rod (14); the two ends of the linkage rod (14) are respectively connected in rotation to the opposite ends of two non-adjacent finger joints among the three corresponding finger joints; and the rotation axis between the linkage rod (14) and the finger joint does not overlap with the rotation axis between adjacent finger joints.

6. A bionic robotic arm according to claim 5, characterized in that: The wrist cross universal joint (3-1) comprises an integrally formed central ring (3-1-1) and four connecting rods; the inner ends of the four connecting rods are all connected to the outer circumferential surface of the central ring (3-1-1); the four connecting rods are respectively two first connecting rods (3-1-2) that are opposite to each other and coaxial, and two second connecting rods (3-1-3) that are opposite to each other and coaxial; The palm support (2) has a rope threading hole at its inner end; the rope threading hole is aligned with the center hole on the center ring (3-1-1); the support knuckle (12) has a clearance hole connected to the internal space of the bionic finger; the finger drive mechanism (5) comprises a steering gear mounting platform (5-1) and a plurality of single-finger drive components; the steering gear mounting platform (5-1) is fixed on the arm support (1); the number of single-finger drive components is equal to and corresponds to the number of bionic fingers; the single-finger drive components comprises a finger drive steering gear (5-2), a winch (5-3) and a rope (5-4 ); the finger-driven servo (5-2) is mounted on the servo mounting platform (5-1); the capstan (5-3) is fixed on the output shaft of the finger-driven servo (5-2); one end of the rope (5-4) is fixed on the capstan (5-3); the rope (5-4) passes through the center hole of the center ring (3-1-1), the rope threading hole on the palm support (2), the clearance hole on the support knuckle (12) of the corresponding bionic finger, and the inner cavity of each movable knuckle (13), and the outer end of the rope (5-4) is fixed to the movable knuckle (13) at the end of the corresponding bionic finger.

7. A bionic robotic arm according to claim 6, characterized in that: Both ends of the center hole of the center ring (3-1-1) are provided with circular arc chamfers.

8. The bionic mechanical arm according to claim 4, characterized in that: An outer shell structure is sleeved and fixed on the outer side of the arm support (1); the palm base (2) and the outer shell structure of the arm support (1) are sealed and connected via an elastic structure.

9. A method for driving a bionic robot arm as claimed in claim 4, characterized in that: When it is necessary to drive the palm support (2) to flip forward or backward, the two direct-acting motors (4-4) are controlled to extend or shorten synchronously, thereby driving the palm support (2) to rotate around the rotation axis of the first rotation pair; When it is necessary to drive the palm support (2) to swing left and right, one of the direct-acting motors (4-4) is synchronously controlled to extend and the other direct-acting motor (4-4) is controlled to shorten, thereby driving the palm support (2) to rotate around the rotation axis of the second rotation pair; During the movement of the palm support (2), the wrist joint dual-axis motion detection flexible sensor collects light intensity signals and resistance signals in real time, and determines the rotation direction and rotation angle of the palm support (2) based on the measured light intensity signals and resistance signals; When it is necessary to drive part or all of the bionic fingers to move, the finger driving mechanism (5) drives the corresponding bionic fingers to bend or straighten.

Images