Self-adaptive tactile sensor for reconstructing three-dimensional shape of object and manufacturing method of self-adaptive tactile sensor

Abstract

The invention discloses a self-adaptive tactile sensor structure and a manufacturing method and application thereof, the self-adaptive tactile sensor structure comprises a three-dimensional neural network and a pressure sensor, the three-dimensional neural network is used for qualitative analysis, and the pressure sensor is used for quantitative analysis. The three-dimensional neural network is composed of multiple layers of simulation neural networks, basic units are simulation neurons, and the simulation neurons are manufactured by simulating the specific response of a human skin tactile receptor and regulating and controlling an activation threshold value through a silk-screen printing process. Furthermore, the simulation neural networks with different activation thresholds are integrated into a three-dimensional neural network through a stack printing process, and the reconstruction of the three-dimensional shape of the object can be realized by applying continuously changing pressure.

Description

Technical Field

[0001] The present invention belongs to a flexible sensing device, specifically an adaptive tactile sensor that imitates the tactile receptors of the human skin, and its structure, manufacturing method and application are introduced. Background Art

[0002] At present, flexible electronic skin is developing towards integration and array. Traditional electronic skin based on flexible pressure sensors faces challenges such as device miniaturization, single function, complex sensing information, and difficult extraction circuit. How to integrate high density, multi-dimensional information, comprehensively extract information and facilitate recognition and processing by modern computer systems in a limited space is the key to promoting the practical application of flexible electronic skin in life. Therefore, new flexible electronic skin urgently needs to break through traditional limitations in sensing principle, device miniaturization and manufacturing process, so as to achieve high-density sensing integration, multi-dimensional complex information integration and highly reliable operation. Summary of the Invention

[0003] The purpose of the present invention is to provide a manufacturing method and application of an adaptive tactile sensor integrating a three-dimensional neural network and a pressure sensor. The three-dimensional neural network is composed of multiple layers of analog neural networks, and the basic unit is an analog neuron with a controllable activation threshold. The proposed analog neuron is inspired by the tactile receptors of the human skin. Tactile receptors distributed in the human skin can produce specific responses to different pressure stimuli, which endows the skin with the ability to distinguish different tactile stimuli. At the same time, a sensing network is formed based on the widely distributed tactile receptors. These sensing networks form a hierarchical response in depth. Based on this, the human brain can judge the shape, texture and other properties of the object being touched. When the pressure stimulus is lower than the activation threshold, the analog neuron does not respond; when the pressure stimulus is higher than the activation threshold, a step jump occurs. By integrating multiple analog neurons with the same activation threshold into an analog neural network, the analog neural network has consistent response characteristics and can make a unified pressure response to the object surface it covers. Based on this characteristic, a three-dimensional neural network composed of multiple layers of analog neural networks with different activation thresholds is combined with a pressure sensor to form an adaptive tactile sensor, which can realize the three-dimensional reconstruction of the object surface shape through the combination of multiple two-dimensional activation maps.

[0004] The technical problems solved by the present invention can be realized by adopting the following technical solutions:

[0005] The adaptive tactile sensor includes a three-dimensional neural network and a pressure sensor, which are divided into a packaging layer, a printed electrode layer, a hollowed dielectric layer, and a pressure-sensitive layer. Among them, the three-dimensional neural network is composed of multiple layers of analog neural networks with different activation thresholds, and the basic unit is an analog neuron. There is a locally hollowed dielectric layer between the upper and lower electrodes of a single analog neuron. This dielectric layer has a certain height and is hollowed between the corresponding upper and lower electrodes. When an external pressure acts on its surface, it is squeezed, causing the upper and lower electrodes to contact and achieve conduction; when the pressure is removed, the upper and lower electrodes separate, and the analog neuron returns to the disconnected state. The pressure stimulus that can cause the analog neuron to change its on-off state is defined as the activation threshold. Among them, there are two ways to control the activation threshold of the analog neuron. The first method is to fix the height of the hollowed dielectric layer and increase the height of the electrodes in the hollowed area, so that the relative air height between the upper and lower electrodes decreases. The method of the embodiment is to use the screen printing process to increase the number of printing times to control the electrode height; the other method is to directly regulate the height of the hollowed dielectric layer, thereby controlling the air height between the upper and lower electrodes to achieve the regulation of the activation threshold.

[0006] The structure and activation method of the analog neuron introduced above are universal. Multiple analog neurons can be directly prepared using the screen printing process to form an analog neural network. Each layer of the analog neural network has a unified activation threshold. When pressure distributed in the plane acts on it, the analog neurons included in the area greater than the activation threshold are activated; the analog neurons in the remaining areas remain disconnected, thereby generating an activation map in the two-dimensional plane. When multiple layers of analog neural networks with different activation thresholds are integrated, due to the pressure stimulus applied increasing (or decreasing) over time, the activation map shown also exhibits a hierarchical effect, and thus the three-dimensional shape of the object being touched can be reconstructed. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is the structure and typical response curve of the analog neuron included in the present invention.

[0008] 1. Upper packaging layer, 2. Upper electrode layer, 3. Hollowed dielectric layer, 4. Lower electrode layer, 5. Lower packaging layer, 6. Hollow cavity

[0009] Figure 2 It is the method for regulating the activation threshold of the analog neuron by controlling the electrode height. 4.1 - 4.4 are the electrode stacks caused by screen printing in the embodiment, so that the air height changes from the initial d to Δd to achieve the purpose of regulation.

[0010] Figure 3 It is the method for regulating the activation threshold of the analog neuron by controlling the height of the hollowed dielectric layer, and directly controls the air height between the upper and lower electrode layers through the height d of the hollowed dielectric layer.

[0011] Figure 4To simulate the planar structure of a neural network.

[0012] Figure 5 To simulate two lead methods of each analog neuron in the neural network. The upper two figures show the individual lead method for each analog neuron; the lower two figures show the parallel interleaved lead method with spaced upper and lower electrodes. 7 is the upper electrode layer, 8 is the lower electrode layer, forming a sandwich structure of upper electrode layer / hollow dielectric layer / lower electrode layer.

[0013] Figure 6 Shows the overall structure and local details of the adaptive tactile sensor. 9 is the three-dimensional neural network, 9.1 - 9.3 are the three-layer analog neural networks of the embodiments, and 10 is the pressure sensor.

[0014] Figure 7 Is the manufacturing flowchart of the adaptive tactile sensor integrating the three-layer analog neural network of the embodiment. Among them, in order to regulate the activation thresholds of different layers of the analog neural network, each layer of the analog neural network needs to be screen-printed multiple times to regulate the electrode height.

[0015] Figure 8 Shows the two-dimensional activation map and three-dimensional shape reconstruction results of the embodiment.

Claims

1. An adaptive tactile sensor, characterized in that: It consists of a three-dimensional neural network and a pressure sensor. The three-dimensional neural network is used for qualitative analysis, and the pressure sensor is used for quantitative analysis. The three-dimensional neural network is composed of multiple layers of analog neural networks, and the basic unit is an analog neuron. It is fabricated by regulating the activation threshold through a silk-screen printing process by imitating the specific response of the tactile receptors in the human skin.

2. The adaptive tactile sensor according to claim 1, characterized in that: The analog neuron consists of a packaging layer, a printed electrode layer, and a hollowed-out dielectric layer. There is a locally hollowed-out dielectric layer between the upper and lower electrode layers of a single analog neuron. This dielectric layer has a certain height, is hollowed out between the corresponding upper and lower electrodes, and is filled with air.

3. An adaptive tactile sensor according to claim 1, characterized in that: The proposed analog neuron has a controllable activation threshold, which is defined as the pressure stimulus that can cause the analog neuron to change its on-off state. Among them, there are two ways to control the activation threshold of the analog neuron. The first method is to fix the height of the hollowed-out dielectric layer and regulate the height of the electrodes in the hollowed-out dielectric layer, so that the air height between the upper and lower electrodes is relatively reduced. In the embodiment, the silk-screen printing process is used to increase the number of printing times to control the electrode height. In fact, this claim is not limited to this method, and other methods such as filling conductive materials can also achieve the purpose of regulating the electrode height, all within the scope of this claim; the other method is to directly regulate the height of the hollowed-out dielectric layer, thereby controlling the air height between the upper and lower electrodes.

4. An adaptive tactile sensor according to claim 1, wherein: The analog neuron has only two states: off and on. Its response curve under alternating pressure shows characteristics similar to the step jump of a digital signal.

5. An adaptive tactile sensor according to claim 1, characterized in that: The three-dimensional neural network is composed of multiple layers of analog neural networks with different activation thresholds, and the analog neurons included in a single layer of analog neural network have the same activation threshold.

6. The self - adaptive tactile sensor according to claim 1, characterized in that: The pressure sensor and the three-dimensional neural network form an adaptive tactile sensor, which can monitor the whole process of pressure stimulation and obtain the curve of pressure change. Different from the quasi-digital signal provided by the analog neuron, the pressure sensor provides an analog signal to provide more refined information as an auxiliary reference.

7. An adaptive tactile sensor according to claim 1, wherein: In the field of flexible sensing, compared with the commonly used silk-screen printing process, the unique structure of the proposed adaptive tactile sensor requires its manufacturing process to include two key processes and their combination: preparing a two-dimensional electrode pattern on the packaging layer through silk-screen printing; realizing interlayer interconnection by hollowing out and filling conductive materials.

8. An adaptive tactile sensor according to claim 1, wherein: This adaptive tactile sensor is different from other tactile sensors. It can directly reconstruct the three-dimensional shape of the object being touched by touch and provide the applied pressure curve to further achieve spatio-temporal analysis.

Images