Manipulator fingertip pressure recognition device and method

Abstract

The invention relates to the technical field of pressure collection, in particular to a manipulator fingertip pressure recognition device and method. The device comprises a contact unit and a processing circuit unit, the contact unit comprises a light resistance shell and a contact circuit board; the contact circuit board comprises photodiodes and light emitting diodes in an annular array. The photoelectric coupling effect is combined, pressure signal collection and transmission are separated, and the device has the advantages of being resistant to electromagnetic interference, not affected by temperature and the like. The method comprises the steps of collecting a voltage signal collected by the photodiode and decoupling pressure by adopting a two-layer graph convolutional network. According to the invention, the directional decoupling is carried out on the pressure signal through the graph convolutional network, so that the device can accurately decompose the pressure information, and a finer recognition capability is provided for three-dimensional pressure induction.

Description

Technical Field

[0001] The present invention relates to the technical field of pressure collection, and in particular to a robot fingertip pressure recognition device and method. Background Art

[0002] The development of robotic arms has made great strides in recent years, with applications ranging from industrial automation to prosthetics and human-robot interaction. The development of advanced robotic arms has greatly increased the demand for accurate and reliable tactile sensing systems that can detect forces applied in three dimensions. Pressure sensing is at the heart of tactile feedback systems that enable robots to perform delicate tasks such as grasping, manipulation, and object handling. Current approaches to integrating pressure sensors into robotic systems often result in bulky, rigid components that lack the flexibility and adaptability required for efficient operation. Summary of the invention

[0003] The invention discloses a device and method for identifying fingertip pressure of a robot, which can improve the robot's ability to provide higher accuracy and real-time tactile feedback in tasks such as grasping, manipulating and fine operation.

[0004] To achieve the above-mentioned object, on the one hand, a robot fingertip pressure recognition device is provided, the device comprising: a contact unit and a processing circuit unit;

[0005] The contact unit comprises a flexible photoresistance housing in the shape of a cover and a contact circuit board installed below the flexible photoresistance housing, a plurality of photodiodes in a circular array are arranged above the contact circuit board, a light emitting diode installed on a contact motor board is arranged at the center of the photodiode, and a connecting line is installed below the contact circuit board;

[0006] The processing circuit unit includes an interface module connected to the connection line, a channel selection module for selecting a signal interface, and a signal processing module.

[0007] Furthermore, the flexible light-resistance housing 1 is made of silicone material and is provided with four light-resistance columns therein.

[0008] Furthermore, the number of the photodiodes is 8.

[0009] Furthermore, the channel selection module is a CD74HC4051 chip, and the signal processing module is a STM32 signal processing chip.

[0010] In order to achieve the above purpose, on the other hand, a method for identifying fingertip pressure of a robot is provided, and the specific method is as follows:

[0011] Construct a functional relationship between voltage and vertical pressure;

[0012] Collecting the voltage signal collected by the photodiode;

[0013] A two-layer graph convolutional network is used to decouple pressure.

[0014] Furthermore, the functional relationship between voltage and vertical pressure is as follows:

[0015] p=kv

[0016] Among them, p is the pressure, k is the coefficient, and v is the voltage value.

[0017] Furthermore, there are 8 photodiodes, and the collected voltage signal is V = [v1, v2, ..., v8] T .

[0018] Furthermore, a two-layer graph convolutional network is used to decouple pressure. The specific method is as follows:

[0019] The input of the two-layer graph convolutional network is the voltage signal V = [v1, v2, ..., v8] T ;

[0020] The output is a pressure decoupled signal, defined as P = [p1, p2, p3] T They represent the X, Y, and Z directions of pressure respectively.

[0021] Furthermore, the two-layer graph convolutional network decouples the pressure, and the graph convolution structure with residual is:

[0022]

[0023] in, It is the node feature matrix of the second layer, and the voltage obtained by the photodiode is used in the first layer.

[0024] are the weights of the first and second layers, which will be learned through training; σ is an activation function; represents the normalized adjacency matrix; by applying the convolution operation layer by layer, the information from the adjacent ports is aggregated to estimate the three-dimensional pressure vector;

[0025] In training, the mean square error is used as the loss function:

[0026]

[0027] Where N is the number of samples, is the predicted pressure vector, P i is the actual pressure vector; the Adam optimizer is used for efficient training, and the weights are updated through back propagation to obtain the relationship between the pressure vector and the voltage signal:

[0028]

[0029] in, Represents the relationship between photodiodes, which are set according to actual measurement analysis:

[0030]

[0031] Due to the adoption of the above technical solution, the present invention has the following beneficial effects:

[0032] 1. The present invention decouples the pressure signal in a directional manner through a graph convolutional network, so that the device can accurately decompose the pressure information and provide a more sophisticated recognition capability for three-dimensional pressure sensing.

[0033] 2. The flexible photoresistive shell made of silicone material has good adaptability and sensitivity, can cope with complex contact situations, and is especially suitable for pressure recognition at the tip of a mechanical finger.

[0034] 3. The CD74HC4051 signal multiplexer chip is used to enable the device to collect voltage signals from multiple photodiodes at the same time, enhancing the system's comprehensive recognition ability of pressure size and direction. In addition, the number of sensors can be expanded by only adding receiving ports and contacts, and multiple sensors can be used in parallel, such as collecting the pressure of 5 fingertips of the manipulator at the same time.

[0035] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings of the present invention are as follows.

[0037] Figure 1 This is a schematic diagram of the structure of the robotic fingertip pressure recognition device.

[0038] Figure 2 Schematic diagram of the flexible photoresist housing structure.

[0039] Figure 3 A top view of the contact circuit board.

[0040] Figure 4 This is a flow chart of the robot fingertip pressure recognition method.

[0041] In the figure: 1. Flexible photoresistor housing; 2. Contact unit; 3. Processing circuit unit; 2.1. Photodiode; 2.2. Contact circuit board; 2.3. Light-emitting diode; 2.4. Connecting wire; 3.1. Interface module; 3.2. Channel selection module; 3.1. Signal processing module. DETAILED DESCRIPTION

[0042] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0043] Embodiment 1:

[0044] A mechanical fingertip pressure recognition device, such as Figure 1 , Figure 2 and Figure 3 As shown, it includes: a contact unit 2 and a processing circuit unit 3;

[0045] The contact unit 2 comprises a cover-shaped flexible photoresistance housing 1 and a contact circuit board 2.2 installed below the flexible photoresistance housing 1, a plurality of photodiodes 2.1 in a circular array are arranged above the contact circuit board 2.2, a light-emitting diode 2.3 installed on the contact motor board 2.2 is arranged at the center of the photodiode 2.1, and a connecting wire 2.4 is installed below the contact circuit board 2.2;

[0046] The processing circuit unit 3 includes an interface module connected to the connection line 2.4, a channel selection module for selecting a signal interface, and a signal processing module.

[0047] The flexible light-resistance housing 1 is made of silicone material and has four light-resistance columns arranged inside.

[0048] There are eight photodiodes.

[0049] The channel selection module is a CD74HC4051 chip that collects the voltage values ​​of 8 photodiodes on the contacts, and the signal processing module is an STM32 signal processing chip that calculates and analyzes the size and direction of the pressure.

[0050] Its working principle is: the photodiode emits light, which is reflected from the inside of the flexible photoresistance housing to the photosensitive diode, which converts the received light signal into a voltage signal. When the flexible photoresistance housing receives pressure, its shape changes. This change in shape changes the optical path of the reflected light, causing the intensity of the light signal received by the photodiode to change, and finally converts it into an electrical signal related to the size and direction of the pressure. The use of a flexible photoresistance housing ensures the sensitivity and adaptability of the device when under pressure, and can effectively reduce the interference of the sensor caused by external pressure changes.

[0051] Embodiment 2:

[0052] A method for identifying fingertip pressure of a robot, such as Figure 4 As shown, the specific steps are as follows:

[0053] S1. Build a functional relationship between voltage and vertical pressure.

[0054] S2, collecting the voltage signal collected by the photodiode.

[0055] S3. Use a two-layer graph convolutional network to decouple pressure.

[0056] In this embodiment, when the flexible photoresistance housing receives pressure, the deformation of the housing causes a change in the light reflection path, thereby affecting the light intensity received by the photodiode. Based on the changes in light intensity of different photodiodes, the system can calculate the magnitude and direction of the pressure. The signal analysis is not limited to the magnitude of the pressure, but can also accurately identify the direction of the pressure based on the different angles of information collected by the photodiode array.

[0057] In order to effectively decouple the pressure component in the three-dimensional direction, the graph convolutional network GCN is used to analyze the pressure direction. The voltage signal of the photodiode is used as input and is defined as V = [v1, v2, ..., v8] T The pressure vector is used as output and is defined as P = [p1, p2, p3] T . Learning and training are performed through a two-layer graph convolutional network model. The structure of the graph convolutional network is:

[0058]

[0059] in, It is the node feature matrix of the second layer, and the voltage obtained by the photodiode in the first layer of the matrix. are the weights of the first and second layers, which will be learned through training. σ is an activation function (such as ReLU). The training process and accuracy of the network are further optimized by introducing residual connections, so as to better decouple the directional information of three-dimensional pressure and enhance the accuracy and stability of pressure recognition.

[0060] The normalized adjacency matrix is ​​used to represent the relationship between photodiodes. Since the photodiodes are evenly distributed on the PCB, the light intensity received by the photodiodes has a related impact, which is analyzed based on actual measurements.

[0061]

[0062] The model can learn information from adjacent ports to accurately estimate the three-dimensional pressure vector. By using the mean square error MSE as the loss function:

[0063]

[0064] Where N is the number of samples, is the predicted pressure vector, P iis the actual pressure vector. Combined with the Adam optimizer, the network is efficiently trained, and the mapping relationship between the pressure vector and the voltage signal is finally obtained:

[0065]

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A mechanical fingertip pressure recognition device, characterized in that: The device comprises: a contact unit (2) and a processing circuit unit (3); The contact unit (2) comprises a cover-shaped flexible light-resistance housing (1) and a contact circuit board (2.2) installed below the flexible light-resistance housing (1); a plurality of photodiodes (2.1) in a ring array are arranged above the contact circuit board (2.2); a light-emitting diode (2.3) installed on a contact motor board (2.2) is arranged at the center of the photodiode (2.1); and a connecting wire (2.4) is installed below the contact circuit board (2.2); The processing circuit unit (3) comprises an interface module (3.1) connected to the connection line (2.4), a channel selection module (3.2) for selecting a signal interface, and a signal processing module (3.3).

2. The robot fingertip pressure recognition device according to claim 1, characterized in that: The flexible light-resistance housing (1) is made of a silicone material and is provided with four light-resistance columns.

3. The robot fingertip pressure recognition device according to claim 1, characterized in that: There are eight photodiodes.

4. The robot fingertip pressure recognition device according to claim 1, characterized in that: The channel selection module is a CD74HC4051 chip, and the signal processing module is an STM32 signal processing chip.

5. The method for identifying fingertip pressure of a robot according to any one of claims 1 to 4, characterized in that: The specific method is as follows: Construct a functional relationship between voltage and vertical pressure; Collecting the voltage signal collected by the photodiode; A two-layer graph convolutional network is used to decouple pressure.

6. The method for identifying fingertip pressure of a robot as claimed in claim 5, characterized in that: The functional relationship between voltage and vertical pressure is as follows: p=kv Among them, p is the pressure, k is the coefficient, and v is the voltage value.

7. The method for identifying fingertip pressure of a robot as claimed in claim 5, characterized in that: There are 8 photodiodes, and the collected voltage signal is V = [v1, v2, ..., v8] T .

8. The method for identifying fingertip pressure of a robot as claimed in claim 7, characterized in that: A two-layer graph convolutional network is used to decouple pressure. The specific method is as follows: The input of the two-layer graph convolutional network is the voltage signal V = [v1, v2, ..., v8] T ; The output is a pressure decoupled signal, defined as P = [p1, p2, p3] T They represent the X, Y, and Z directions of pressure respectively.

9. The method for identifying fingertip pressure of a robot as claimed in claim 8, characterized in that: The two-layer graph convolution network decouples the pressure, and the graph convolution structure with residual is: in, It is the node feature matrix of the second layer, and the voltage obtained by the photodiode is used in the first layer. are the weights of the first and second layers, which will be learned through training; σ is an activation function; represents the normalized adjacency matrix; by applying the convolution operation layer by layer, the information from the adjacent ports is aggregated to estimate the three-dimensional pressure vector; In training, the mean square error is used as the loss function: Where N is the number of samples, is the predicted pressure vector, P i is the actual pressure vector; the Adam optimizer is used for efficient training, and the weights are updated through back propagation to obtain the relationship between the pressure vector and the voltage signal: in, Represents the relationship between photodiodes, which are set according to actual measurement analysis:

Images