A bionic multi-finger underwater manipulator device, control system, grasping strategy and target recognition method

Abstract

The application relates to a kind of bionic multi-finger underwater manipulator device, control system, grabbing strategy and target identification method, manipulator is mainly composed of flange and seven-link mechanism underactuated manipulator finger, manipulator finger includes base, triangular piece, fingertip piece and slider, two angles of triangular piece are connected with fingertip piece and slider respectively through connecting rod mechanism, the third angle is connected with base through trapezoidal connecting piece, slider is driven by threaded linear long shaft motor, magnetic touch sensor is arranged in fingertip, and magnetic touch sensor receives fingertip touch information through fingertip receptor unit made of magnetic powder silica gel mixture. The device is controlled by the control system containing microprocessor, and the corresponding adaptive force grabbing strategy is formulated to adapt to objects of different shapes and sizes, and underwater objects are classified and identified through a deep learning algorithm. Compared with the prior art, the application has the advantages of high sensitivity, anti-interference and the like, and has the ability of tactile sensing identification and adaptive grabbing.

Description

Technical Field

[0001] This invention relates to a multi-fingered robotic hand, and more particularly to a biomimetic multi-fingered underwater robotic hand device, control system, grasping strategy, and target recognition method. Background Technology

[0002] With the increasing demand for underwater operations, underwater robots have been widely used in deep-sea exploration, scientific research, resource exploration, and environmental monitoring. In these applications, the design and control of manipulators have become key technologies. In the underwater environment, the working conditions of manipulators are more complex, affected by various factors such as water pressure, current, and object characteristics. Therefore, manipulators with rigid designs can provide stable grasping force and reliable manipulation performance. This design typically includes multi-degree-of-freedom joint structures that can simulate the movement of human fingers, thereby achieving effective grasping and manipulation of objects. At the same time, the robustness of rigid manipulators gives them better durability and reliability when facing high water pressure and strong currents.

[0003] In recent years, with the rapid development of artificial intelligence and sensing technology, intelligent robotic arms incorporating tactile perception have gradually become a research hotspot. By integrating tactile sensors, robotic arms can monitor the applied force in real time during the grasping process and adjust their grasping strategy based on feedback information, achieving adaptive grasping. This approach significantly improves the operational performance of robotic arms in complex underwater environments, enabling them to better perform various tasks, such as grasping and recognizing objects.

[0004] However, current underwater manipulators still have some shortcomings. For example, their drive structures are relatively complex, resulting in high maintenance costs; underwater tactile sensors have low sensitivity, weak anti-interference capabilities, and insufficient durability; the integration, control, and recognition algorithms of tactile sensors have not been fully optimized; and the manipulators cannot precisely adjust the gripping angle and force according to the characteristics of objects, are not sensitive enough when facing slippage, and are not accurate enough in recognizing objects. Therefore, developing a lightweight, underactuated underwater bionic manipulator that combines efficient tactile perception and intelligent control to meet the needs of flexible configuration in various working modes has become an urgent need to improve the efficiency and intelligence of underwater operations. Summary of the Invention

[0005] The purpose of this invention is to overcome the lack of highly sensitive and interference-resistant tactile perception and recognition capabilities, as well as adaptive grasping capabilities in existing robotic arms. It provides a novel, lightweight, biomimetic multi-finger underwater robotic arm device, control system, grasping strategy, and target recognition method that has tactile perception function, flexible flange size adjustment, and is integrated into the end effector of various underwater robots or underwater equipment.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] According to one aspect of the present invention, a biomimetic multi-finger underwater manipulator is provided, comprising a fixed flange and a plurality of mechanical fingers connected to the side of the fixed flange;

[0008] The mechanical finger includes a base, a triangular component, a fingertip component, and a slider. The three corners of the triangular component are connected to the fingertip component, the slider, and the base respectively through a linkage mechanism. The point where the triangular component is connected to the base through the linkage mechanism is a fixed rotation point. The slider is used to push one corner of the triangular component to rotate around the fixed rotation point through linear motion, thereby causing the fingertip component to swing.

[0009] The slider is driven by a threaded linear long-axis motor. The fingertip component is equipped with a magnetic tactile sensor, which receives tactile information from the fingertip through a fingertip sensor unit, which is a mixture of magnetic powder and silicone.

[0010] Furthermore, the mechanical finger is a seven-bar linkage mechanism. The base is fixedly connected to the fixed flange. The seven-bar linkage mechanism includes a driving long link, a triangular piece, a first finger joint link, a first finger key, a second finger pad, a third finger joint link, and a fingertip link. One end of the triangular piece is connected to the slider through the driving long link, and the other end is connected to the first finger key. The middle part is connected to the base through a trapezoidal connector.

[0011] The first finger key is connected to the triangular piece at one end and to the second finger pad at the other end; the second finger pad is connected to the first finger key at one end and to the fingertip connecting rod at the other end, and is connected to the first knuckle connecting rod in the middle; the first knuckle connecting rod is connected to the middle of the triangular piece at one end and to the third knuckle connecting rod at the other end, and is connected to the second finger pad at the middle end; the third knuckle connecting rod is connected to the fingertip connecting rod at one end and to the end of the first knuckle connecting rod at the other end; the fingertip connecting rod connects the second finger pad and the third knuckle connecting rod.

[0012] The slider is located on the base and has the freedom to reciprocate along the axis of the mechanical finger, with the fingertip embedded in the surface of the fingertip link.

[0013] Furthermore, the mechanical finger also includes the fingertip, which is fixedly connected to the first phalanx link.

[0014] Furthermore, the fingertip component includes a fingertip base, a magnetic tactile sensor, and an isolation bracket connected in sequence. The fingertip base has a notch, and the fingertip component is inserted into the fingertip link through the notch on the fingertip base.

[0015] Furthermore, the isolation bracket has a groove in which the fingertip sensor unit is located, and the isolation bracket separates the magnetic tactile sensor and the fingertip sensor unit.

[0016] Furthermore, a motor compartment is provided on the side of the fixed flange, a threaded linear long shaft motor is fixed in the motor compartment, and the motor shaft extends out from the motor compartment and is connected to the slider. An AB dual-phase Hall encoder is provided behind the threaded linear long shaft motor.

[0017] Furthermore, the motor housing includes a motor base and a housing. The motor base is fixedly connected to a fixed flange, and the housing is connected to the motor base via a tenon and mortise structure. The AB dual-phase Hall encoder is connected to the threaded linear long-shaft motor, and the motor base and housing seal the AB dual-phase Hall encoder and the threaded linear long-shaft motor.

[0018] According to another aspect of the present invention, a control system for a biomimetic multi-fingered underwater robotic hand device is provided, characterized in that it includes a microprocessor, a voltage regulator module, a crystal oscillator circuit, a reset circuit, a motor drive module, a flash memory, tactile sensors, and indicator lights; the microprocessor uses an STM32F103C8T6 as the main control chip, and is equipped with a voltage regulator module, a crystal oscillator circuit, and a reset circuit, and uses an SWD interface as a programmer, and is connected to the motor drive module through an I / O port; the motor drive module generates PWM to drive the threaded linear long-axis motor 2 of each robotic finger; the microprocessor is also connected to the flash memory through an SPI bus interface, and connects to or reads the magnetic tactile sensor 152 corresponding to each robotic finger through an IIC bus interface and address selection; the control system displays status information through indicator lights.

[0019] According to another aspect of the present invention, a grasping strategy for a biomimetic multi-fingered underwater manipulator is provided, characterized by comprising the following steps:

[0020] The magnetic tactile sensor monitors the changes in friction between the fingertip and the object in real time during the grasping process;

[0021] When the system detects a change in friction, it determines whether the change in friction exceeds the slip detection threshold.

[0022] When the change in friction exceeds the slip detection threshold, the adjustment amount of the gripping force of the robotic arm device is calculated by combining the target gripping force and using the gripping force adjustment formula;

[0023] Control the linear long-axis motor of the thread to adjust the friction force until the change in friction force does not exceed the slip detection threshold.

[0024] According to another aspect of the present invention, a target recognition method for a biomimetic multi-fingered underwater manipulator is provided, characterized by comprising the following steps:

[0025] Feature vectors are extracted from the contact force data collected by the magnetic tactile sensor;

[0026] The output of the convolutional layer is obtained from the feature vector using a convolutional neural network deep learning model;

[0027] The output of the convolutional layer is downsampled to obtain the output of the pooling operation;

[0028] The extracted features are mapped to the output layer through a series of fully connected layers, and the number of neurons in the output layer corresponds to the number of categories of the target object.

[0029] The output is activated to generate the classification result C of the target object.

[0030] Compared with the prior art, the present invention has the following advantages:

[0031] 1. The mechanical finger of this invention adopts a seven-link underactuated mechanism design, which simplifies the transmission scheme and uses a single driver to control multiple joints. It can accurately adjust the grasping force and angle according to the characteristics of the object. The multi-finger coordinated grasping and single-finger grasping methods make it more reliable. The rigid linkage manipulator has better durability and reliability when facing high water pressure and strong water flow. In addition, the magnetic tactile sensor set at the fingertip of this invention has the advantages of high sensitivity, strong anti-interference ability and high durability, which can provide accurate multi-dimensional force perception in complex environments and improve its perception ability.

[0032] 2. The invention proposes a strategy for adjusting the gripping force in real time based on slip detection data. This strategy can detect slip in real time using a magnetic tactile sensor and adaptively adjust the gripping force based on the slip data, thereby improving gripping stability.

[0033] 3. The underwater bionic three-finger manipulator with tactile sensing function designed in this invention uses a deep learning algorithm to extract the features required for grasping objects. By evaluating the grasping force applied to different objects, the system can infer the shape and hardness of the objects, thereby achieving accurate target identification and improving its intelligent operation level. Attached Figure Description

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

[0035] Figure 2 This is a schematic diagram of the mechanical finger structure of the present invention;

[0036] Figure 3 This is an exploded view of the mechanical finger structure of the present invention;

[0037] Figure 4 This is an exploded view of the fixed flange structure of the present invention;

[0038] Figure 5 This is an exploded view of the fingertip component of the present invention;

[0039] Figure 6 This is a schematic diagram of the control system framework of the present invention;

[0040] Figure 7 This is a control block diagram of the adaptive gripping force adjustment strategy of the present invention;

[0041] Figure 8 This is a schematic diagram of the deep learning algorithm flow of the present invention;

[0042] In the attached diagram: 1-Fixed flange; 2-Threaded linear long-shaft motor; 3-Base; 4-Slider cover; 5-Slider; 6-Drive long connecting rod; 7-Trapezoidal connector; 8-Triangular piece; 9-Finger pad; 10-First finger joint connecting rod; 11-First finger key; 12-Second finger pad; 13-Third finger joint connecting rod; 14-Fingertip connecting rod; 15-Fingertip piece; 151-Fingertip base; 152-Magnetic tactile sensor; 153-Isolation bracket; 16-Motor compartment; 17-AB dual-phase Hall encoder; 18-Hexagonal nut. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0044] This invention provides an underwater bionic three-finger manipulator with tactile sensing capabilities. The main body of the manipulator includes a fixed flange 1 and a manipulator located on the side of the fixed flange 1. The manipulator adopts an underactuated anthropomorphic hand design and is a seven-bar linkage mechanism, including a triangular member 8. Through the motion transmission of the linkages, it achieves functions such as three-finger coordinated grasping, single-finger grasping, and soil digging. Specifically, the first finger joint linkage 10 is embedded in a fingertip 9, and the fingertip linkage 14 is embedded in a fingertip member 15. In particular, the fingertip member 15 embeds a magnetic tactile sensor 152 and a fingertip sensor unit. The fingertip sensor unit is manufactured by adding magnetic powder to silicone and then molding it, and is isolated from the magnetic tactile sensor 152 by a bracket. A microprocessor is installed in the waterproof sealed chamber of various underwater robots or underwater equipment carrying this manipulator to control its movement. The manipulator is driven by three threaded linear long-axis motors 2 and is equipped with an AB dual-phase Hall encoder 17 for accurate attitude reading. The robotic arm achieves multiple functions such as three-finger coordinated grasping, single-finger grasping, and soil removal and sand digging through ingenious linkage motion transmission.

[0045] The working principle of this invention is as follows: An underwater bionic three-finger robotic hand with tactile sensing function is fixed to the end of an underwater robot or underwater equipment. The robotic hand transmits motion through a clever linkage, precisely adjusting the gripping force and angle according to the object's characteristics to achieve collaborative gripping with all three fingers. It can also independently grasp slender or spherical objects using a single finger. For objects hidden in soft sand or mud, the fingertips can be used for preliminary digging and grasping. Furthermore, the robotic hand can detect slippage in real time using a magnetic tactile sensor 152 and adaptively adjust the gripping force based on the slippage data. When the sensor detects a slight slippage of the object at the fingertip, the system gradually increases the gripping force through feedback control to prevent the object from falling. Finally, the robotic hand collects contact force data in real time during the gripping process using the magnetic tactile sensor 152, and uses this data to identify and classify the target object.

[0046] like Figure 1 As shown, the threaded linear long-shaft motor 2 is fixed to the motor compartment 16 on the side of the fixed flange 1. The motor compartment 16 consists of a motor base and a housing, which are connected using mortise and tenon joints to seal the threaded linear long-shaft motor 2. The threaded linear long-shaft motor 2 is connected to the slider 5 via a hexagonal nut 18. The hexagonal nut 18 is installed inside the slider 5, and a slider cover 4 is provided above the slider 5 to fit the hexagonal nut 18 and prevent it from falling out. Figure 2 As shown, the rotational motion of the threaded linear long-axis motor 2 is converted into the vertical up-and-down motion of the slider 5.

[0047] like Figure 3 As shown, slider 5 reciprocates up and down on base 3, and trapezoidal connector 7 is fixed below base 3 by positioning pins. Slider 5 is connected to drive long connecting rod 6 by pins, and the other side of drive long connecting rod 6 is connected to one end of triangular piece 8. The top of triangular piece 8 is connected to trapezoidal connector 7 and first knuckle connecting rod 10 by screws. First knuckle connecting rod 10 is fixed to fingertip 9 by positioning pins. The other end of triangular piece 8 is fixed to one side of first finger key 11 by screws. The other side of first finger key 11 is fixed to one side of second fingertip 12 by positioning pins, and the middle hole of second fingertip 12 is fixed to the middle hole of first knuckle connecting rod 10 by positioning pins. The other side of first knuckle connecting rod 10 is connected to third knuckle connecting rod 13 by positioning pins. The other side of third knuckle connecting rod 13 is connected to one positioning hole of fingertip connecting rod 14 by positioning pins. The other side of second fingertip 12 is connected to another positioning hole of fingertip connecting rod 14 by positioning pins. A fingertip base 151 is embedded in the surface of the fingertip link 14, and the fingertip base 151 is the first layer of the fingertip component 15. For example... Figure 5As shown, the second layer of the fingertip component 15 houses the magnetic tactile sensor 152, the third layer is the isolation bracket 153, and the fourth layer is the fingertip sensor unit, which is a hybrid magnetic powder silicone body. The fingertip sensor unit is made by AB silicone molding.

[0048] In a preferred embodiment, a fingertip 9 is fixedly connected to the first knuckle link 10. The fingertip 9 is used for single-finger grasping. When the finger is bent to a certain extent, the fingertip 9 will contact the trapezoidal connector 7, thereby fixing the first knuckle link 10 of the finger. The triangular member 8 continues to drive the first finger key 11, the second fingertip 12, the third knuckle link 13, the fingertip link 14 and the fingertip member 15 to rotate further, thereby realizing single-finger bending.

[0049] like Figure 2 As shown, when the threaded linear long shaft motor 2 drives the slider 5 to move backward on the base 3, the slider 5 drives the driving long connecting rod 6 to move backward. Under the pull of the driving long connecting rod 6, the triangular piece 8 rotates counterclockwise around the connection between the triangular piece 8 and the trapezoidal connecting piece 7. At this time, the connection between the triangular piece 8 and the driving long connecting rod 6 moves backward, and the connection between the triangular piece 8 and the first finger key 11 moves forward. The first finger key 11 moves forward under the push of the triangular piece 8. The fingertip 9, the first knuckle link 10, the first finger key 11, the second fingertip 12, the third knuckle link 13, the fingertip link 14, and the fingertip piece 15 bend inward as a whole. When the bending movement reaches a certain degree, the fingertip 9 will contact the trapezoidal connector 7, thereby restricting the further movement of the first knuckle link 10 of the finger. The triangular piece 8 continues to push the first finger key 11. The first finger key 11 drives the second fingertip 12, the third knuckle link 13, the fingertip link 14, and the fingertip piece 15 to rotate further around the connection between the second fingertip 12 and the first knuckle link 10, thereby realizing single-finger bending.

[0050] like Figure 4 As shown, a microprocessor is located at the bottom of the fixed flange 1, used to control the threaded linear long-axis motor 2 to drive multiple joints and adjust the movement of each joint. An AB dual-phase Hall encoder 17 is mounted above the threaded linear long-axis motor 2 for precise reading of the robot's posture. The microprocessor is connected to a voltage regulator module (PW2126, TPS54302), indicator lights, a reset circuit, a crystal oscillator circuit, a programming interface, and a motor drive module (TB6612FNG) via I / O ports. It is connected to three magnetic tactile sensors 152 via an IIC bus and to a Flash memory (AT45DB64) via SPI. The microprocessor generates three PWM waves via a timer. These PWM waves, after passing through the motor drive module, control the threaded linear long-axis motor 2 to achieve adjustable speed rotation.

[0051] As mentioned earlier, the fingertip component 15 embeds a magnetic tactile sensor 152. The magnetic tactile sensor 152 consists of four ultra-small digital three-dimensional Hall sensors MLX90393 and an FPC interface. It senses force or contact by detecting changes in the magnetic field of the fingertip sensor unit. It typically comprises a magnetic field sensing element (such as a Hall effect sensor) and an embedded magnetic source. When an external force or contact is applied, the position or magnetic field strength of the magnetic source changes. The magnetic field sensor converts this into an electrical signal, which, after processing, provides information on the magnitude and direction of the force.

[0052] In particular, the present invention also includes a grasping perception and recognition algorithm equipped on the robotic arm. Through the collaborative work of the microprocessor and the magnetic tactile sensor 152, the grasping force can be adjusted in real time according to the slip detection data, thereby adapting to objects of different shapes and sizes and effectively grasping and recognizing underwater objects.

[0053] The grasping strategy proposed in this invention can detect slippage in real time using a magnetic tactile sensor 152 and adaptively adjust the grasping force based on the slippage data. When the sensor detects a slight slippage of an object on the fingertip, the system gradually increases the grasping force through feedback control to prevent the object from falling off. Slippage is sensed by monitoring changes in the magnetic field and pressure fluctuations, and the increment and rate of the grasping force are adjusted according to the degree of slippage to ensure that the grasping force is moderate and does not damage the object. In addition, the force adjustment is further optimized using friction coefficient estimation, making the grasping process more stable. Through this closed-loop feedback mechanism, the robot can flexibly respond to the grasping needs of different objects in an underwater environment. The process includes the following steps:

[0054] S1. The change in friction between the fingertip and the object during the grasping process is monitored in real time by the magnetic tactile sensor 152. The formula for calculating the friction is:

[0055] F f =μ·F n (1)

[0056] Where μ is the coefficient of friction (which depends on the contact state between the object's surface material and the tactile sensor), F n Normal force (the pressure exerted on an object by the fingertip).

[0057] S2. When the system detects a change in friction, it determines whether the change in friction exceeds the slip detection threshold. The slip detection threshold can be set using the following formula:

[0058] S th =α·μ·F n (2)

[0059] Among them, S th The sliding threshold is α, which is a preset constant, usually between 1.1 and 1.5, used to set a safety margin.

[0060] S3. When the change in friction exceeds the slip detection threshold, the gripping force adjustment of the robotic arm is calculated by combining the target gripping force and using the gripping force adjustment formula. The gripping force adjustment can be based on slip detection feedback.

[0061] F g (t)=F g (t-1)+k·(F target -F g (t-1)) (3)

[0062] Among them, F g (t) represents the grasping force F at the current moment. g (t-1) represents the grasping force at the previous moment, F target The target gripping force is (based on slip detection and friction estimation), and k is the adjustment rate constant (between 0 and 1, controlling the increment).

[0063] S4. Control the threaded linear long shaft motor 2 to adjust the friction force until the change in friction force does not exceed the slip detection threshold.

[0064] This invention also relates to a target recognition method. Based on the constant grasping angle of the controlled robot, a deep learning algorithm is used for target recognition and detection. A magnetic tactile sensor 15-2 collects contact force data in real time during the grasping process, and uses this data to identify and classify target objects. The system uses a deep learning model to analyze sensor data, extracting the characteristics required for grasping, such as the required grasping force and the object's physical properties. By evaluating the grasping force applied to different objects, the system can infer the object's shape, hardness, etc., thereby achieving accurate target recognition. The method includes the following steps:

[0065] S1. Extract feature vectors from the contact force data collected by the magnetic tactile sensor 152. The feature vector X extracted by the data processing module consists of the following:

[0066] X = [F] g ,ΔF g ,T,V,…] (4)

[0067] Where, ΔF g Let T be the rate of change of the grasping force, T be the contact time, and V be the grasping speed;

[0068] S2. The convolutional layer output is obtained from the feature vector using a convolutional neural network (CNN) deep learning model. The feature extraction process can be represented as follows:

[0069] Z=σ(W*X+b) (5)

[0070] Where Z is the output of the convolutional layer, W is the convolutional kernel, b is the bias, and σ is the activation function;

[0071] S3. Downsample the output of the convolutional layer to reduce data dimensionality and computational cost, obtaining the output of the pooling operation, which is represented as:

[0072] P = pool(Z) (6)

[0073] Here, pool is a pooling operation (such as max pooling or average pooling);

[0074] S4. The extracted features are mapped to the output layer through a series of fully connected layers. The number of neurons in the output layer corresponds to the number of categories of the target object. The output calculation can be expressed as:

[0075] O=σ(W f ·P+b f (7)

[0076] Among them W f It is the weight matrix of the fully connected layer, b f It is the bias, and 0 is the final output;

[0077] S5. Activate the output to generate the classification result C of the target object, usually using the Softmax function:

[0078]

[0079] Among them, C i It is the probability of the target category, k is the total number of categories, O i It is the value of the i-th neuron in the output layer.

[0080] The specific embodiments of the present invention are described below through examples:

[0081] Example 1

[0082] like Figures 1-5 As shown, this embodiment provides a biomimetic multi-finger underwater manipulator with embodied sensing and flexible multi-mode configuration, including a fixed flange 1 and a three-finger manipulator located on the side of the fixed flange 1. The manipulator adopts an underactuated design method and is a seven-bar linkage structure, including a triangular member 8, with the first finger joint link 10 fixedly connected to the fingertip 9. The fingertip base 151 of the fingertip component 15 is embedded in the surface of the fingertip link 14, and the fingertip base 151 is the first layer of the fingertip component 15. The second layer of the fingertip component 15 houses the magnetic tactile sensor 152, the third layer is the isolation bracket 153, and the fourth layer is the fingertip sensor unit. The fingertip sensor unit is a mixed magnetic powder silicone body, which is made by AB silicone molding. A threaded linear long-axis motor 2 is fixed to the motor housing 16 on the side of the fixed flange 1. The motor housing 16 is connected using the tenon and mortise principle to seal the threaded linear long-axis motor 2.

[0083] The control system of the robotic arm is centered on a microprocessor, which is installed in a waterproof, sealed compartment within the underwater robot or equipment carrying the robotic arm to control its movement. For example... Figure 6 As shown, the microprocessor uses an STM32F103C8T6 as the main control chip, equipped with a power supply regulator module (TPS54302), a crystal oscillator circuit (8MHz), and a reset circuit (power-on reset). It uses an SWD interface as a programmer and connects to a motor drive module (TB6612FNG) via I / O ports. The motor drive module generates PWM to drive the threaded linear long-axis motor 2 and the threaded linear long-axis motors corresponding to the second and third mechanical fingers. The main control chip connects to a flash memory (AT45DB641E) via an SPI bus interface and to a magnetic tactile sensor 15-2 (MLX90393) via an IIC bus interface. Furthermore, address selection allows for reading the magnetic tactile sensors 152 corresponding to the second and third mechanical fingers. The control system also displays device status information via LED indicators.

[0084] Example 2

[0085] Building upon Example 1, this example further tests the three-finger coordinated grasping capability of the underwater bionic robotic hand. The underwater bionic three-finger robotic hand designed in this study employs a precise seven-bar linkage mechanism, enabling the grasping of various objects by individually and simultaneously controlling the movement angles of the three robotic fingers. This robotic hand can intelligently adjust the grasping force and angle based on the characteristics of the target object (such as shape and size). For example, when facing milk bottles, small metal balls (such as ping-pong balls), rubber jellyfish, and rubber shrimp, the robotic hand can adjust the grasping angles of the three robotic fingers according to the object's characteristics, ensuring a uniform distribution of grasping force and preventing the object from slipping due to concentrated force.

[0086] Example 3

[0087] like Figure 1 and Figure 2 As shown, based on Embodiment 1 or Embodiment 2, the single-finger design of the robotic arm provides flexible soil-scraping and grasping functions for hidden objects in soft sandy environments. The robotic arm embeds a magnetic tactile sensor 152 within its fingertip to perceive the physical properties of the sand in real time. During operation, the robotic arm first uses its fingertip to scrape away the soil, clearing the layer of sand covering the target object. For example, when searching for a small ceramic vase buried in fine sand, the robotic arm can easily scrape away the sand and obtain the hardness information of the sand using the built-in tactile sensor 152 to optimize the subsequent grasping strategy. Furthermore, the precise angle control of the multi-link system ensures that the robotic arm can grasp slender or small spherical objects, such as small agate blocks or ceramic blocks, with a single finger.

[0088] Example 4

[0089] like Figure 7 As shown, based on Embodiment 1, Embodiment 2, or Embodiment 3, to ensure object stability during the grasping process, the magnetic tactile sensor 152 equipped on the robotic arm can monitor possible slippage in real time. When grasping heavy objects (such as metal blocks), after the system detects a change in the frictional force applied to the object, the control algorithm automatically calculates and gradually increases the grasping force to prevent the object from slipping. In specific implementation, the robotic arm dynamically adjusts the grasping force based on changes in frictional force, contact time, and the dynamic characteristics of the object. For example, when slippage of an object is detected, the robotic arm will optimize the grasping force in real time through a closed-loop feedback mechanism to ensure flexible response to different grasping needs in the underwater environment.

[0090] Example 5

[0091] like Figure 8 As shown, based on Embodiment 1, Embodiment 2, or Embodiment 3, the robotic arm classifies objects using real-time collected contact force data. During the grasping process, the robotic arm analyzes and extracts features of different objects, such as small rubber underwater organisms, small metal balls, milk bottles, small agate blocks, pebbles, keys, and ballpoint pens. Through a deep learning model, the system can classify and identify target objects based on feature vectors such as changes in grasping force, contact time, and grasping speed. In the implementation process, a convolutional neural network (CNN) is used to process the collected multidimensional data, thereby accurately inferring the shape, hardness, and other physical properties of the objects. This algorithm not only improves the robotic arm's ability to identify various underwater objects but also provides data support for subsequent autonomous decision-making, helping to efficiently complete tasks in complex underwater environments.

[0092] In the specific implementation of the above embodiments, the technical features can be combined in any non-contradictory way. For the sake of brevity, not all possible combinations of the above technical features are described. However, as long as the combination of these technical features is not contradictory, it should be considered to be within the scope of this specification.

[0093] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A biomimetic multi-fingered underwater robotic hand device, characterized in that, Includes a fixed flange (1) and multiple mechanical fingers connected to the side of the fixed flange (1); The mechanical finger includes a base (3), a trapezoidal connector (7), a triangular piece (8), a fingertip piece (15), and a slider (5). Two corners of the triangular piece (8) are connected to the fingertip piece (15) and the slider (5) respectively through a linkage mechanism. The third corner of the triangular piece (8) is connected to the base (3) through the trapezoidal connector (7), and the connection point is a fixed rotation point. The slider (5) is used to push one corner of the triangular piece (8) to rotate around the fixed rotation point through linear motion, thereby causing the fingertip piece (15) to swing. The slider (5) is driven by a threaded linear long-axis motor (2). The fingertip component (15) is equipped with a magnetic tactile sensor (152). The magnetic tactile sensor (152) receives fingertip tactile information through a fingertip sensor unit. The fingertip sensor unit is a magnetic powder silicone mixture. The mechanical finger is a seven-bar linkage mechanism. The base (3) is fixedly connected to the fixed flange (1). The seven-bar linkage mechanism includes a driving long link (6), a triangular piece (8), a first finger joint link (10), a first finger key (11), a second finger pad (12), a third finger joint link (13), and a fingertip link (14). One end of the triangular piece (8) is connected to the slider (5) through the driving long link (6), and the other end is connected to the first finger key (11). The middle part is connected to the base (3) through a trapezoidal connector (7). One end of the first finger key (11) is connected to the triangular piece (8), and the other end is connected to the second finger pad (12); one end of the second finger pad (12) is connected to the first finger key (11), and the other end is connected to the fingertip connecting rod (14), and the middle end is connected to the first knuckle connecting rod (10); one end of the first knuckle connecting rod (10) is connected to the middle of the triangular piece (8), and the other end is connected to the third knuckle connecting rod (13), and the middle end is connected to the second finger pad (12); one end of the third knuckle connecting rod (13) is connected to the fingertip connecting rod (14), and the other end is connected to the end of the first knuckle connecting rod (10); the fingertip connecting rod (14) is connected to the second finger pad (12) and the third knuckle connecting rod (13); the slider (5) is located on the base (3) and has the freedom to reciprocate along the axis of the mechanical finger; the fingertip piece (15) is embedded in the surface of the fingertip connecting rod (14); The mechanical finger also includes a fingertip (9), which is fixedly connected to the first phalanx link (10); The fingertip (9) is used for single-finger grasping. When the finger is bent to a certain degree, the fingertip (9) contacts the trapezoidal connector (7), thereby fixing the first finger joint link (10) of the finger. The triangular part (8) continues to drive the first finger key (11), the second fingertip (12), the third finger joint link (13), the fingertip link (14) and the fingertip part (15) to rotate further, thereby realizing single-finger bending. The magnetic tactile sensor (152) includes multiple three-dimensional Hall sensors that sense force or contact through changes in the magnetic field of the fingertip sensor unit. The grasping strategy of the biomimetic multi-fingered underwater manipulator includes the following steps: The changes in friction between the fingertip and the object during the grasping process are monitored in real time using a magnetic tactile sensor (152); When the system detects a change in friction, it determines whether the change in friction exceeds the slip detection threshold. When the change in friction exceeds the slip detection threshold, the adjustment amount of the gripping force of the robotic arm device is calculated by combining the target gripping force and using the gripping force adjustment formula; Control the threaded linear long shaft motor (2) and adjust the friction until the change in friction does not exceed the slip detection threshold.

2. The biomimetic multi-fingered underwater robotic hand device according to claim 1, characterized in that, The fingertip component (15) includes a fingertip base (151) and an isolation bracket (153). The fingertip base (151) has a notch, and the fingertip component (15) is inserted into the fingertip connecting rod (14) through the notch on the fingertip base (151). The isolation bracket (153) has a groove, and the fingertip sensor unit is located in the groove. The isolation bracket (153) separates the magnetic tactile sensor (152) and the fingertip sensor unit.

3. The biomimetic multi-fingered underwater manipulator device according to claim 1, characterized in that, The fixed flange (1) has a motor compartment (16) on its side. The threaded linear long shaft motor (2) is fixed in the motor compartment (16), and the motor shaft extends out from the motor compartment (16) and is connected to the slider (5). An AB dual-phase Hall encoder (17) is provided behind the threaded linear long shaft motor (2).

4. The biomimetic multi-fingered underwater manipulator device according to claim 3, characterized in that, The motor housing (16) includes a motor base and a housing. The motor base is fixedly connected to the fixed flange (1). The housing is connected to the motor base through a tenon and mortise structure. The AB dual-phase Hall encoder (17) is connected to the threaded linear long shaft motor (2). The motor base and the housing seal the AB dual-phase Hall encoder (17) and the threaded linear long shaft motor (2).

5. A control system for a biomimetic multi-fingered underwater manipulator according to any one of claims 1-4, characterized in that, The system includes a microprocessor, a voltage regulator module, a crystal oscillator circuit, a reset circuit, a motor drive module, a flash memory, a tactile sensor, and indicator lights. The microprocessor uses an STM32F103C8T6 as the main control chip, and is equipped with a voltage regulator module, a crystal oscillator circuit, and a reset circuit. It also has an SWD interface and is connected to the motor drive module via an I / O port. The motor drive module generates PWM to drive the threaded linear long-axis motor (2) of each mechanical finger. The microprocessor is also connected to the flash memory via an SPI bus interface and connects to or reads the magnetic tactile sensor (152) corresponding to each mechanical finger via an IIC bus interface and address selection. The control system displays status information via indicator lights.

6. A target recognition method for a biomimetic multi-fingered underwater manipulator according to any one of claims 1-4, characterized in that, Includes the following steps: Feature vectors are extracted from the contact force data collected by the magnetic tactile sensor (152); The output of the convolutional layer is obtained from the feature vector using a convolutional neural network deep learning model; The output of the convolutional layer is downsampled to obtain the output of the pooling operation; The extracted features are mapped to the output layer through a series of fully connected layers, and the number of neurons in the output layer corresponds to the number of categories of the target object. The output is activated to generate the classification result C of the target object.

Images