Integrated fiber bragg grating fingertip tactile sensor and its encapsulation manufacturing method

Abstract

The application discloses an integrated fiber Bragg grating fingertip tactile sensor and a packaging manufacturing method thereof, comprising a fingertip base, a fingertip connecting piece, a fiber array and a packaging layer; the fingertip connecting piece is fixedly installed on one side of the fingertip base and is used for connecting an external machine finger structure; the fiber array comprises a plurality of optical fibers, and a plurality of fiber Bragg grating sensing units are arranged on each optical fiber; a plurality of groups of positioning bosses corresponding to the plurality of optical fibers are fixedly arranged on the surface of the fingertip base, a fiber penetrating hole is formed in the middle of each positioning boss, so that each optical fiber penetrates through the fiber penetrating hole of one group of positioning bosses; and the packaging layer is used for covering the fingertip base and the fiber array. The application improves packaging manufacturing reliability and integrated installation reliability, improves the spatial positioning precision and packaging consistency of the fiber array, reduces the risk of optical fiber breakage, bending and deviation in the manufacturing process, and is thus favorable for improving the flexible linear object posture estimation and the robot operation control effect.

Description

Technical Field

[0001] This invention relates to the field of tactile sensor technology, specifically to an integrated fiber Bragg grating fingertip tactile sensor and its packaging and manufacturing method. Background Technology

[0002] Existing sensors are primarily designed for single-point force, distributed pressure, three-dimensional force, temperature, or slippage detection. However, flexible linear objects such as cables, ropes, conduits, flexible tubes, wire harnesses, and surgical sutures are prone to continuous deformation, self-occlusion, and localized bending during robotic gripping. Traditional vision methods struggle to stably observe the gripped area, and compact, camera-free tactile sensors still fall short in directly sensing the contact posture of flexible linear objects. Related research indicates that posture estimation of flexible linear objects relies on tactile sensing technologies capable of directly sensing fingertip contact deformation.

[0003] Fiber Bragg gratings (FBGs) are a commonly used type of fiber optic sensing element. When an external contact load causes deformation of the encapsulated elastomer, the resulting strain is transmitted to the FBG, causing a change in the center wavelength or spectral morphology of its reflection peak, thereby enabling the sensing of contact force, deformation, or contact position. Because bare FBGs are made of quartz glass, their mechanical strength is limited, making them susceptible to damage during impact, bending, shearing, or repeated contact. Therefore, current technology typically encapsulates FBGs in flexible elastomers. Polydimethylsiloxane (PDMS) and other silicone rubber-like materials, due to their flexibility, elasticity, and moldability, have been used for flexible encapsulation of FBGs.

[0004] While existing fiber Bragg grating tactile sensors have been used in scenarios such as contact force detection, 3D force measurement, temperature compensation, and slip sensing in robot fingers, some solutions still employ independent flexible sheets, externally attached sensing units, or finger sleeve structures. These solutions lack an integrated mechanical connection between the sensor body and the robot's fingertip base, often relying on adhesives or external fixing structures for installation. This makes them difficult to directly adapt to space-constrained areas such as robot fingertips and the ends of industrial grippers. For example, existing square-hole flexible FBG tactile sensors require attachment to the robot's finger surface after fabrication. While they enable tactile sensing and monitoring, their installation reliability and replaceability are still affected by the external attachment method.

[0005] Furthermore, fiber Bragg gratings are formed from silica optical fibers, and bare fibers are prone to breakage or performance degradation during bending, stretching, shearing, potting, curing, and demolding. In the current embedded FBG sensor packaging process, fiber breakage reduces manufacturing yield; existing solutions struggle to achieve integrated positioning and packaging, and the packaging layer is prone to detachment; most existing FBG tactile sensors also employ discrete structures, requiring complex external fixing structures, making direct integration into confined spaces such as the dexterous fingertips of robots difficult.

[0006] Therefore, existing technologies have not fully solved the problems of difficult integration and installation, insufficient packaging and manufacturing reliability, easy breakage of optical fibers, easy detachment of packaging layers, and difficulty in stably obtaining contact posture information of flexible linear objects in the application of fiber Bragg grating tactile sensors in robot fingertip applications. It is still necessary to propose an integrated fiber Bragg grating fingertip tactile sensor and its packaging and manufacturing method. Summary of the Invention

[0007] In view of the problems existing in the prior art, the purpose of this invention is to provide an integrated fiber Bragg grating fingertip tactile sensor and its packaging and manufacturing method.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An integrated fiber Bragg grating fingertip tactile sensor includes a fingertip base, a fingertip connector, a fiber array, and an encapsulation layer; The fingertip connector is fixedly installed on one side of the fingertip base, and the fingertip connector is used to connect to an external robotic finger structure; The fiber array includes multiple optical fibers, and each optical fiber is provided with several fiber Bragg grating sensing units. The surface of the fingertip base is fixedly provided with multiple sets of positioning protrusions corresponding to multiple optical fibers. The positioning protrusions have fiber-passing holes in the middle, so that each optical fiber passes through the fiber-passing holes of a set of positioning protrusions. The encapsulation layer is used to cover the fingertip base and the fiber array.

[0009] In some embodiments, the encapsulation layer is formed by mixing polydimethylsiloxane liquid and a curing agent at a mass ratio of 10:1 and then curing them in a molding die.

[0010] In some embodiments, the surface of the fingertip base is further provided with a plurality of fixed protrusions, which are used to form a mechanical fit with the encapsulation layer.

[0011] In some embodiments, each optical fiber is provided with multiple fiber Bragg grating sensing units. Each fiber Bragg grating sensing unit has a corresponding center reflection wavelength, and the center wavelengths of different fiber Bragg grating sensing units on each fiber are staggered from each other.

[0012] In some embodiments, the fiber array includes two optical fibers, and two sets of corresponding positioning protrusions are fixedly provided on the surface of the fingertip base. The two optical fibers are arranged in parallel through the two sets of positioning protrusions, and four fiber Bragg grating sensing units are provided on each optical fiber.

[0013] Another aspect of the present invention provides a packaging and manufacturing method for the above-mentioned integrated fiber Bragg grating fingertip tactile sensor, comprising the following steps: S1. Complete the manufacturing of the fingertip base, fingertip connector and multiple optical fibers respectively. The structure and size of the fingertip connector are matched with the corresponding machine finger structure. The optical fibers contain fiber Bragg grating sensing units that meet the requirements. S2. Fix the fingertip base to the fingertip connector, and pass multiple optical fibers through the fiber holes on multiple sets of positioning protrusions so that each optical fiber reaches the preset position. S3. The overall structure processed in step S2 is sent into the preset molding mold and assembled. S4. Place the assembled molding die into the preset packaging alignment device, which is used to fix the molding die; S5. Prepare a mixture for forming the encapsulation layer; S6. Inject the mixture into the mold cavity of the molding die; S7. The mixture is injected into the mold and kept in a stable environment for curing to obtain the encapsulation layer; S8. After curing, release the sealing alignment device from the molding mold and demold to complete the manufacturing of the integrated fiber Bragg grating fingertip tactile sensor.

[0014] In some embodiments, in step S1, the fingertip base and the fingertip connector are manufactured by 3D printing; After the optical fibers are fabricated, they are connected to a fiber Bragg grating demodulator to detect whether the initial reflection peak of the fiber Bragg grating sensing unit of each optical fiber meets the requirements.

[0015] In some embodiments, in step S1, when the manufacturing of the fingertip base is completed, each set of positioning bosses of the fingertip base has one positioning boss located at one edge of the fingertip base. In step S2, when each optical fiber reaches the preset position, both ends of it extend to the area outside the fingertip base, and one end serves as the lead-out end, which is provided with a connector structure for connecting to the fiber grating demodulator, while the other end serves as the reserved end and passes through the positioning boss located at the edge of the fingertip base. In step S4, the encapsulation alignment device also clamps both ends of each optical fiber, so that each optical fiber is in a pre-tensioned state. In step S8, after demolding, the reserved end of the optical fiber is bonded and fixed to the corresponding positioning boss, and the optical fiber located outside the positioning boss is cut off to ensure that the cut surface is flat, thus completing the manufacturing of the integrated fiber Bragg grating fingertip tactile sensor.

[0016] In some embodiments, in step S5, specifically: Add the polydimethylsiloxane liquid and the curing agent to a clean container at a mass ratio of 10:1, stir thoroughly to ensure uniform dispersion of the curing agent, and obtain a mixture. After stirring, place the mixture in a vacuum environment for 15 to 20 minutes to degas, allowing the bubbles in the mixture to rise and burst.

[0017] In some embodiments, after step S8, the method further includes the following step: S9. Connect the fiber end of the fabricated integrated fiber Bragg grating fingertip tactile sensor to a fiber grating demodulator and check whether the reflection peak of each fiber meets the requirements when it is stationary; and check whether the reflection peak of each fiber meets the requirements when it is in operation by pressing the surface of the encapsulation layer on the corresponding side of the fiber.

[0018] Compared with the prior art, the present invention has the following significant advantages: The integrated fiber Bragg grating fingertip tactile sensor and its packaging manufacturing method provided by this invention differ from the existing scheme of attaching independent flexible sensor sheets to the surface of robot fingers. Instead, the fiber array is directly arranged and packaged on the fingertip base to form an integrated fingertip component. This improves the spatial positioning accuracy and packaging consistency of the fiber array, enhances the reliability of packaging manufacturing and integration installation, and reduces the risk of fiber breakage, bending and displacement during manufacturing. In turn, it helps to improve the attitude estimation of flexible linear objects and the operation control effect of robots. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the integrated fiber Bragg grating fingertip tactile sensor provided by the present invention; Figure 2 This is a schematic diagram showing the integrated fiber Bragg grating fingertip tactile sensor provided by the present invention in conjunction with a molding die and packaging alignment device during the manufacturing stage. Figure 3This is a schematic diagram of the original FBG reflectance spectrum collected in a specific embodiment of the present invention; Figure 4 This is a schematic diagram for obtaining the attitude coordinates of a flexible linear object. Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the following description, in conjunction with the accompanying drawings and specific embodiments, further explains how this invention is implemented.

[0021] The present invention provides an integrated fiber Bragg grating fingertip tactile sensor, comprising a fingertip base 1, a fingertip connector 2, an optical fiber array, and an encapsulation layer 3.

[0022] In this embodiment, the fingertip base 1 serves as a rigid support structure for the sensor, and the fingertip connector 2 is fixedly mounted on one side of the fingertip base 1. The fingertip connector 2 is used to connect to an external robotic finger structure. The aforementioned robotic finger structure may be, for example, a robot finger joint, a dexterous hand end effector, a parallel gripper, or other gripper.

[0023] The fingertip base 1 and the fingertip connector 2 are preferably manufactured by 3D printing, but can also be manufactured by injection molding, machining or other molding methods. The shape of the fingertip connector 2 can be designed according to the installation space at the end of the robot's finger, and a connection part that mates with the fingertip base 1 is reserved.

[0024] The fingertip connector 2 can adopt screw holes, slots, pin holes, plugs, threaded parts, clamping grooves, or a one-piece molded connection structure. For different robot fingers or grippers, only the shape, hole position, or connection size of the fingertip connector 2 needs to be changed to adapt to different installation interfaces, without changing the basic structure of the fingertip base 1 and the encapsulation layer 3.

[0025] The fiber array includes multiple optical fibers 4, and each optical fiber 4 is provided with several fiber Bragg grating sensing units 41; the surface of the fingertip base 1 is fixedly provided with multiple sets of positioning protrusions 11 corresponding to the multiple optical fibers 4, and the positioning protrusions 11 have fiber-passing holes in the middle, so that each optical fiber 4 passes through the fiber-passing holes of a set of positioning protrusions 11.

[0026] Preferably, the positioning boss 11 is a small block-shaped boss, and the diameter of the fiber optic hole is slightly larger than the outer diameter of the optical fiber 4, allowing the optical fiber to pass through smoothly while limiting the lateral movement of the optical fiber. The opening of the fiber optic hole should avoid sharp edges as much as possible, and chamfers or rounded corners can be set to reduce the risk of the optical fiber coating being scratched during the fiber insertion process. By setting the positioning boss 11, the spatial path of the optical fiber 4 can be restricted, so that each fiber Bragg grating sensing unit 41 is located at a predetermined sensing position. Furthermore, during the subsequent encapsulation layer 3 filling process, it can prevent the optical fiber 4 from drifting with the liquid flow and reduce local bending or loosening of the optical fiber 4. Overall, the positioning boss 11 plays a guiding, limiting and protecting role for the optical fiber 4.

[0027] Preferably, in Figure 1 In the illustrated embodiment, the fiber array includes two optical fibers 4. Two sets of positioning protrusions 11 are fixedly disposed on the surface of the fingertip base 1. The two optical fibers 4 are arranged parallel to each other via the two sets of positioning protrusions 11. Each optical fiber 4 is provided with four fiber Bragg grating sensing units 41. Preferably, both optical fibers 4 are located inside the encapsulation layer 3, approximately 2 mm from the tactile surface, and close to the neutral surface of the encapsulation layer 3. By employing multiple fiber Bragg grating sensing units 41 to cover the tactile area, richer spatial contact information can be obtained than with a single-point force sensor, facilitating the characterization of the contact position, contact angle, and local posture of flexible linear objects. Other numbers and arrangements of optical fibers 4 can also be used depending on actual needs. Furthermore, one end of the optical fiber 4 serves as a lead-out end, which can be equipped with an existing connector structure for connection to an existing fiber Bragg grating demodulator during operation. The fiber Bragg grating demodulator is then connected to a computer or robot control system.

[0028] Furthermore, each fiber Bragg grating sensing unit 41 has a corresponding center reflection wavelength, and the center wavelengths of different fiber Bragg grating sensing units 41 on each fiber 4 are staggered. Additionally, the center wavelength of each fiber Bragg grating sensing unit 41 is within the operating wavelength range of the fiber Bragg grating demodulator used, so that the demodulator can simultaneously identify each reflection peak. For example, if a fiber Bragg grating demodulator with a wavelength range of 1530 nm to 1570 nm is used, the center wavelength of each fiber Bragg grating sensing unit 41 can be designed according to this range, ensuring sufficient spacing between different reflection peaks.

[0029] Encapsulation layer 3 is used to cover the fingertip base 1 and the fiber array. In this embodiment, encapsulation layer 3 is formed by mixing polydimethylsiloxane liquid and curing agent at a mass ratio of 10:1 and then curing them in molding mold 5. In other embodiments, the polydimethylsiloxane material can also be replaced with silicone rubber, flexible polyurethane, thermoplastic elastomer, or other elastic polymer materials with flexibility, elastic recovery capability, and moldability. The alternative material should be able to cover the fiber array and produce repeatable elastic deformation under external contact loads.

[0030] One outer end face of the encapsulation layer 3 (the outer end face corresponding to the side where the optical fiber 4 is located) can serve as the tactile area. The preferred size of the tactile area is 30mm × 11.25mm, which can cover the main contact area of ​​flexible linear objects such as cables, ropes, flexible tubes, conduits, and surgical sutures on the fingertips. The encapsulation layer 3 acts as a flexible force transmission medium to transmit the surface deformation caused by external contact loads to the internal fiber Bragg grating sensing unit 41; and it also acts as a protective layer for the optical fiber 4, preventing the bare optical fiber 4 from directly bearing impact, shearing, and friction; the encapsulation layer 3 can adapt to the local contact pattern of the flexible linear object through elastic deformation.

[0031] Preferably, the surface of the fingertip base 1 is further provided with a plurality of fixed bosses 12, which are used to form a mechanical fit with the encapsulation layer 3. The fixed bosses 12 can be a plurality of small bosses spaced apart, or they can be columnar, toothed, inverted, annular, or roughened protrusion structures.

[0032] The fixed boss 12 can be covered or partially covered by the encapsulation layer 3. By setting the fixed boss 12 to form a mechanical fit with the encapsulation layer 3, the encapsulation layer 3 and the fingertip base 1 not only rely on the adhesion of the material surface, but also have anti-peeling and anti-slip effects through the fixed boss 12, reducing the risk of the encapsulation layer 3 falling off during robot operation. The fixed boss 12 can also be used to cooperate with the inner wall or positioning surface of the molding die 5 to form a limiting relationship, thereby helping to limit the thickness and shape of the encapsulation layer 3.

[0033] Another aspect of the present invention provides a packaging and manufacturing method for the above-mentioned integrated fiber Bragg grating fingertip tactile sensor, comprising the following steps: S1. Complete the manufacturing of fingertip base 1, fingertip connector 2 and multiple optical fibers 4 respectively. The structure and size of the fingertip connector 2 are matched with the corresponding machine finger structure. The optical fiber 4 contains a fiber Bragg grating sensing unit 41 that meets the requirements.

[0034] The fingertip base 1 and the fingertip connector 2 are manufactured using 3D printing. After 3D printing, the supporting structure and burrs should be removed, and any residue inside the fiber optic hole of the positioning boss 11 should be cleaned to ensure that the optical fiber 4 can pass through smoothly. If necessary, a fine needle, micro-drill, or airflow can be used to clear the fiber optic hole; the opening of the fiber optic hole 21 can also be slightly chamfered to reduce the risk of the optical fiber being scratched during installation.

[0035] Preferably, during the manufacturing of the fingertip base 1, each set of positioning bosses 11 of the fingertip base 1 has one positioning boss 11 located at one edge of the fingertip base 1, corresponding to... Figure 1 The two positioning bosses 11 on the far left of the embodiment shown.

[0036] In addition, after the fiber 4 is fabricated, it can be connected to a fiber Bragg grating demodulator to check whether the initial reflection peak of the fiber Bragg grating sensing unit 41 of each fiber 4 meets the requirements. According to the test results, if the reflection peak of each fiber Bragg grating sensing unit 41 is complete, the peak shape is normal, and the baseline is stable, the next step can be carried out; if there is a missing reflection peak, severe peak shape distortion, or fiber breakage, fiber 4 should be replaced. To facilitate subsequent positioning, marks can be made on the outside of fiber 4 near each fiber Bragg grating sensing unit 41; the marks should avoid damaging the fiber coating and should ensure that the approximate position of each fiber Bragg grating sensing unit 41 can be identified during fiber threading and assembly.

[0037] S2. Fix the fingertip base 1 to the fingertip connector 2, and pass multiple optical fibers 4 through the fiber holes on multiple sets of positioning bosses 11, so that each optical fiber 4 reaches the preset position.

[0038] When threading the fiber, the operation should be slow to avoid sharp bends, twists, knots, or scratches on the fiber 4 through the aperture. Each fiber Bragg grating sensing unit 41 should be positioned at a predetermined sensing location within the tactile area. Preferably, after encapsulation, the fiber Bragg grating sensing unit 41 is located approximately 2 mm from the surface of the tactile area, a position that balances sensing sensitivity and damage resistance. After the fiber 4 is threaded, temporary clamps or low-stress fixing methods can be used to secure the end of the fiber 4 to prevent significant displacement.

[0039] Furthermore, when each optical fiber 4 reaches the preset position, both ends extend to the area outside the fingertip base 1, with one end serving as the lead-out end and equipped with a connector structure for connecting to the fiber optic demodulator, and the other end serving as the reserved end and passing through the positioning boss 11 located at the edge of the fingertip base 1.

[0040] S3. The overall structure processed in step S2 is sent to the preset molding mold 5 and assembled.

[0041] During assembly, the fixing boss 12 of the fingertip base 1 should mate with the inner wall or positioning surface of the forming mold 5 to form a stable limit. After assembly, the following should be checked: whether each optical fiber 4 is still in the predetermined position; whether the forming mold 5 is pressing the optical fiber; whether the injection port and venting channel of the forming mold 5 are unobstructed; and whether the fixing boss 12 mates correctly with the forming mold 5.

[0042] The molding die 5 can be an integral die, an upper and lower mold, a left and right split die, a side-pulling die, or a detachable clamping structure. As long as the die can define the thickness of the encapsulation layer 3, the fiber embedding depth, and the shape of the tactile surface, the encapsulation purpose of this invention can be achieved.

[0043] S4. Place the assembled molding mold 5 into the preset packaging alignment device 6, which is used to fix the molding mold 5.

[0044] like Figure 2 As shown, the encapsulation alignment device 6 also clamps both ends of each optical fiber 4, ensuring that each optical fiber 4 is in a pre-tensioned state. Pre-tensioning does not involve applying a large pulling force to the optical fiber 4, but rather keeping it slightly straight and stable to prevent loosening, floating, local bending, or serpentine bending of the optical fiber 4 during the encapsulation layer 3 infusion and curing process. Furthermore, Figure 2 The upper fiber 4 shown is the aforementioned reserved fiber 4, which needs to be cut off after demolding.

[0045] S5. Prepare a mixture for forming the encapsulation layer 3.

[0046] Specifically, the polydimethylsiloxane liquid and the curing agent are added to a clean container at a mass ratio of 10:1, and stirred thoroughly to ensure the curing agent is evenly dispersed, resulting in a mixture. Vigorous stirring should be avoided during the process to minimize the introduction of air bubbles.

[0047] After stirring, the mixture is placed in a vacuum environment for degassing for 15 to 20 minutes, allowing the bubbles in the mixture to rise and burst. Vacuum degassing can reduce bubbles inside the encapsulation layer 3, prevent bubbles from affecting strain transmission near the fiber Bragg grating sensing unit 41, and also help improve the mechanical consistency of the tactile surface.

[0048] S6. Inject the mixture into the mold cavity of the molding mold 5.

[0049] The degassed mixture can be slowly injected into the mold cavity of the molding mold 5 using a syringe. During injection, it should be slowly advanced along one side of the mold cavity or the predetermined injection port so that the mixture gradually fills the mold cavity. The mixture should completely cover the fiber array and at least cover the side wall or part of the top of the fixing boss 12 so that the cured encapsulation layer 3 and the fixing boss 12 form a mechanical fit.

[0050] Furthermore, rapid impacts to fiber 4 should be avoided during the infusion process to prevent the fiber Bragg grating sensing unit 41 from shifting. After infusion, the area can be left to stand for a period of time to further release any remaining tiny air bubbles.

[0051] S7. The mixture is injected into the completed molding mold 5 and kept in a stable environment for curing to obtain the encapsulation layer 3. Preferably, curing is performed at room temperature for approximately 24 hours. During the curing process, the encapsulation alignment device 6 should continue to fix the molding mold 5 and the end of the optical fiber 4 to prevent mold movement or optical fiber loosening.

[0052] After curing, polydimethylsiloxane changes from a liquid state to an elastic solid state, forming the encapsulation layer 3. At this time, the fiber Bragg grating fiber array 41 is stably encapsulated in the encapsulation layer 3, the fixing boss 12 forms a mechanical fit with the encapsulation layer 3, and the positioning boss 11 maintains the spatial position of the fiber Bragg grating sensing unit 41.

[0053] S8. After curing, release the sealing alignment device 6 from the molding mold 5 and demold to complete the manufacturing of the integrated fiber Bragg grating fingertip tactile sensor.

[0054] When demolding, the operation should be slow to avoid directly pulling the lead end of fiber 4. After demolding, the appearance of the encapsulation layer 3 should be checked to confirm that there are no obvious missing glue, bubbles, cracks, tears, edge lifting or local detachment. The lead end of fiber 4 should also be checked to confirm that there are no breaks, dents or excessive bending.

[0055] After demolding, the reserved end of the optical fiber 4 is bonded and fixed to the corresponding positioning boss 11, and the optical fiber 4 located outside the positioning boss 11 is cut off to ensure that the cut surface is flat, thus completing the manufacturing of the integrated fiber Bragg grating fingertip tactile sensor.

[0056] S9. Connect the fiber 4 end of the integrated fiber Bragg grating fingertip tactile sensor to the fiber grating demodulator and check whether the reflection peak of each fiber 4 meets the requirements when it is stationary; and check whether the reflection peak of each fiber 4 meets the requirements when it is working by pressing the surface of the encapsulation layer 3 on the corresponding side of the fiber 4.

[0057] The testing can include: the presence of each reflection peak, the integrity of the peak shape, whether the center wavelength is within the expected range, the stability of the baseline in a static state, and whether each reflection peak produces a responsive wavelength shift when the tactile surface is slightly pressed. If the reflection peaks of each fiber Bragg grating sensing unit 41 are complete and the baseline is stable, it indicates that no obvious fiber breakage, grating breakage, or severe residual stress abnormality occurred during the packaging process.

[0058] This invention provides an integrated fiber Bragg grating fingertip tactile sensor that can be directly mounted on the fingertips of robots. This allows multiple optical fibers 4, each equipped with a fiber Bragg grating sensing unit 41, to be stably encapsulated within the fingertip base 1 at predetermined positions, depths, and tensions, forming a stable and reliable tactile surface. Simultaneously, it reduces the risks of fiber 4 breakage, fiber 4 misalignment, encapsulation layer 3 detachment, and inconsistent tactile responses, providing a stable spectral acquisition hardware foundation for contact deformation sensing, attitude estimation, and closed-loop robot operation of flexible linear objects. This invention exhibits good structural adaptability and scalability, and can be used in dexterous hands, parallel grippers, industrial grippers, medical robot end effectors, or other devices requiring fingertip tactile sensing.

[0059] After the integrated fiber Bragg grating fingertip tactile sensor is manufactured, the usage process is as follows: First, securely install the fiber optic cable 2 onto the robot finger, dexterous hand end, or industrial gripper end, with the outer surface of the encapsulation layer 3 facing the object to be grasped, and connect each fiber optic cable 4 to the fiber optic grating demodulator through the connector at the lead end. The fiber optic grating demodulator is then connected to a computer or robot control system.

[0060] When the robot's fingers grasp flexible linear objects such as cables, ropes, conduits, flexible tubes, wire bundles, or surgical sutures, the flexible linear objects come into contact with the tactile surface of the encapsulation layer 3. The contact load causes localized compression, bending, or shear deformation of the polydimethylsiloxane encapsulation layer 3. This deformation is transmitted to the internal fiber Bragg grating sensing unit 41, causing different degrees of reflection peak center wavelength shift and spectral shape changes in the fiber Bragg grating sensing unit 41 at different locations.

[0061] The central reflection wavelength of a fiber Bragg grating can be expressed as: λ B =2nT Where, λ B λ is the center reflection wavelength of the fiber Bragg grating, n is the effective refractive index of the fiber core, and T is the grating period. When external contact causes strain in the fiber Bragg grating sensing unit 41, n and T change, leading to λ B A shift occurs. Therefore, by collecting the center wavelength shift, peak width change, or complete spectral shape change of the reflection peak of each fiber Bragg grating sensing unit 41, and combining it with relevant algorithms, the contact deformation information of the flexible linear object on the fingertip tactile surface can be obtained.

[0062] In one specific embodiment, the integrated fiber Bragg grating fingertip tactile sensor (hereinafter referred to as FBG tactile sensor) provided by the present invention is used to realize the tactile posture estimation of a flexible linear object through the following method: In step S1, a flexible wire is held by a robotic finger containing an FBG tactile sensor to collect the raw FBG reflectance spectrum.

[0063] In this embodiment, the FBG tactile sensor uses two FBG optical fibers, with four FBG sensing units mounted on each fiber. Therefore, the acquired raw FBG reflectance spectrum is divided into two channels, each containing four sampling nodes, for a total of eight sampling nodes.

[0064] After the robot's finger grasps a flexible thread, the thread applies a contact force to the surface of the FBG tactile sensor, causing a slight deformation of the FBG optical fiber. The FBG reflection peak undergoes wavelength shift, spectral peak broadening, and peak shape changes with the deformation.

[0065] In step S2, peak detection is performed on the original FBG reflection spectrum to locate the center position of each FBG reflection peak, and a local region of interest with a fixed width is obtained by cropping around the center of each FBG reflection peak.

[0066] In this embodiment, multiple local regions of interest are represented as , The region of interest corresponds to the i-th FBG reflection peak. N represents the number of FBG reflection peaks (8 in this embodiment, corresponding to 8 sampling nodes). W represents the width of each region of interest. W can be set according to the actual situation, as long as the width can preserve information such as peak shape, peak width, and side lobes.

[0067] Within the obtained local region of interest, information such as the peak position, peak shape, peak width, side lobes, and local noise patterns of the FBG reflection peak can be preserved.

[0068] In this embodiment, the acquired raw FBG reflectance spectrum is as follows: Figure 3 As shown in the figure, two channels and eight FBG reflection peaks are visible. Additionally, Figure 2 The small box at the top center represents the region of interest corresponding to one of the FBG reflection peaks. It shows the change of a single FBG reflection peak as the clamping force increases. The percentages on the right represent the clamping force level. In this embodiment, 10%, 30%, 50%, 70%, and 90% correspond to clamping forces of 5N, 15N, 25N, 35N, and 45N, respectively.

[0069] In step S3, each local region of interest is encoded as a graph node feature to obtain an initial node feature matrix.

[0070] In this embodiment, each local region of interest is... Input a shared one-dimensional encoder to obtain the corresponding graph node features, represented as follows: .

[0071] in, This represents a one-dimensional encoder, where d represents the feature dimension of the graph node, which can be 32, 64, or 128.

[0072] Stacking all the graph node features yields the initial node feature matrix, represented as: .

[0073] After this processing, each graph node feature corresponds to an FBG reflection peak, rather than a regular sampling point. This reduces interference from irrelevant spectral regions and ensures a one-to-one correspondence between the graph structure and the actual FBG sensing unit.

[0074] In step S4, a multi-relation graph prior for peak perception is constructed based on the initial node feature matrix.

[0075] The constructed prior representation of the multi-relation graph regarding peak perception is as follows: ; in, , , Let these represent the physical adjacency matrix, the spectral correlation matrix, and the learnable adjacency matrix, respectively. , , and All 1-th order matrix, , , They are respectively , , The corresponding learnable weight scalar, This indicates row-based normalization.

[0076] The physical adjacency matrix is ​​used to describe the structural adjacency relationships of graph node features on the same fiber channel, and can be expressed by the following formula: ; It can be seen that if two graph node features i and j are adjacent on the same fiber channel, the corresponding element is 1; otherwise, it is 0. This relationship provides prior knowledge of the sensor's physical structure.

[0077] The spectral correlation matrix is ​​used to describe the spectral similarity between local regions of interest (ROIs) of different FBG reflection peaks. Preferably, the Pearson correlation coefficient can be used to calculate the similarity between any two ROIs, in which case the spectral correlation matrix can be expressed by the following formula: ; in, Represents the i-th local region of interest The kth element, for The mean of the values ​​can be used. The spectral correlation matrix can be used to model cross-FBG spectral response coupling caused by deformation of the same contact posture or adjacent contacts. In other embodiments, cosine similarity, Euclidean distance kernel function, radial basis function kernel function, or learnable similarity function can also be used to obtain the spectral correlation matrix.

[0078] The learnable adjacency matrix is ​​calculated through the dot product attention between graph node features, and is expressed as: ; in: ; ; and For a learnable projection matrix, For the attention dimension.

[0079] By , , Weighted fusion yields a multi-relationship graph prior A, which contains structural information of the FBG tactile sensor, as well as sample spectral information and data-driven implicit relationships.

[0080] In step S5, deep graph reasoning is performed based on the constructed multi-relation graph prior to obtain the final node feature matrix.

[0081] In this embodiment, deep graph inference is performed through L stacked inference layers, where L can be set to 2 to 6 as needed. For any one of the first L-1 inference layers, local graph inference is first performed using the dynamic K-nearest neighbor algorithm to obtain the locally enhanced node feature matrix. Then, the locally enhanced node feature matrix is ​​input into the graph Transformer self-attention module to calculate the global dependencies across nodes, resulting in the globally enhanced node feature matrix. After deep graph inference through the first L-1 inference layers, the final node feature matrix is ​​obtained through the Lth inference layer.

[0082] In any of the first In each inference layer, When obtaining the locally enhanced node feature matrix, specifically: No. The node feature matrix of each inference layer input is represented as follows: Through the dynamic K-nearest neighbor algorithm, based on The resulting binary mask matrix is ​​represented as follows: ; in, Indicates the first The feature space of each inference layer is the set of the k nearest nodes to node i, where k can be 2 to 4.

[0083] Then, the binary mask matrix is ​​multiplied element-wise with the prior of the multi-relation graph and row-normalized to obtain the propagation weights of that layer. : ; in, This indicates element-wise multiplication; The node feature matrix after local enhancement is obtained by the following formula. : ; in: ; in, All are learnable matrices. For learnable parameters, is the activation function for the Gaussian error linear unit.

[0084] This process allows the graph topology to be both constrained by the prior of a stable structure and dynamically adjusted according to the characteristics of the current sample, thus adapting to different flexible linear materials, diameters, contact postures, and clamping forces.

[0085] Next, in any number of... In each inference layer, When obtaining the globally enhanced node feature matrix, specifically: To capture global dependencies across nodes, a graph Transformer self-attention module is applied to all nodes. The globally enhanced node feature matrix is ​​represented as follows: ; in: ; and All are learnable projection matrices. All are attention dimensions.

[0086] based on The node feature matrix for the next inference layer is obtained through linear projection, residual connection, and layer normalization. .

[0087] Specifically, first Projecting back to the original feature dimension, it is represented as: ; in, It is a learnable matrix.

[0088] Then, using the standard Transformer form for residual connections and layer normalization, it can be represented as follows: ; ; in, Representation layer normalization, feedforward network The GELU activation function is applied independently to each node.

[0089] like Then, the final node feature matrix is ​​obtained in the Lth inference layer. .

[0090] This invention enables the model to gradually integrate multi-relational priors, adaptive local interactions, and global contextual dependencies by stacking multiple such inference layers.

[0091] In step S6, based on the final node feature matrix, graph-level features are obtained through global attention pooling, and then the graph-level features are input into the multilayer perceptron to output the attitude coordinates of the flexible linear object.

[0092] In this embodiment, based on the final node feature matrix ,set up express The i-th row obtains graph-level features through global attention pooling. Represented as: ; in, for The corresponding attention pooling weights can be learned.

[0093] Then, map-level features Input a multilayer perceptron and output the attitude coordinates of the flexible linear object.

[0094] In this embodiment, when the robot finger grips a flexible linear object, the gripping surface is rectangular. After the flexible linear object is gripped, a contact line or approximately a contact line is formed on this area. The flexible linear object crosses the x-direction of the rectangular gripping surface, and the y-coordinate of the intersection point of the two ends of the contact area between the flexible linear object and the gripping surface is used to determine the attitude of the flexible linear object. Figure 4 As shown in the figure, the rectangle represents the clamping surface. and The two ends of the contact area between the flexible linear object and the clamping surface are visible. and The x-coordinate is fixed, and the y-coordinate reflects the attitude of the flexible linear object. This representation can uniquely describe the position and tilt direction of the contact line of the flexible linear object within the gripping surface, and can be directly used for robot gripping attitude correction.

[0095] Therefore, when the multilayer perceptron outputs the attitude coordinates of the flexible linear object, it can be expressed as: ; in, , , , All are learnable parameters. This refers to the predicted attitude coordinates of the flexible linear object.

[0096] In addition, during the training phase, in step S1, raw reflection spectra of the FBG under different flexible linear objects, different contact positions, different contact angles, and different clamping forces are collected as training samples, and the true attitude coordinates are labeled for each collected training sample; after processing in steps S2-S6, the predicted attitude coordinates are output; the mean square error between the predicted attitude coordinates and the true attitude coordinates is used as the training loss to complete the training, and the training loss is expressed as: ; Where M represents the number of training samples, and Let represent the predicted attitude coordinates and the actual attitude coordinates of the m-th sample, respectively.

[0097] After training is completed, tactile posture estimation of flexible linear objects can be achieved through steps S1-S6 in the practical stage.

[0098] This invention directly regresses the attitude coordinates of a flexible linear object from the original FBG reflection spectrum end to end, without the need to explicitly solve for contact force, axial strain, radial strain, or the attitude of the flexible body's end, thereby reducing the propagation of intermediate errors.

[0099] Additionally, based on the output The system determines whether the contact line of the flexible linear object deviates from the target centered position. If the contact line is skewed or offset, the robot correction control module generates planar movement or rotation correction actions based on the deviation, gradually bringing the flexible linear object's posture in the fingertip tactile area closer to the target position. Experimental results show that after introducing this tactile posture estimation, the success rate of the robot's flexible linear object posture correction task increased from 80.1% to 91.7%, and the average final posture error decreased from 2.12 mm to 1.84 mm.

[0100] Finally, it should be noted that the above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An integrated fiber Bragg grating fingertip tactile sensor, characterized in that, Includes fingertip base (1), fingertip connector (2), fiber array and encapsulation layer (3); The fingertip connector (2) is fixedly installed on one side of the fingertip base (1), and the fingertip connector (2) is used to connect to an external machine finger structure; The fiber array includes multiple optical fibers (4), and each optical fiber (4) is provided with several fiber Bragg grating sensing units (41). The surface of the fingertip base (1) is fixedly provided with multiple sets of positioning bosses (11) corresponding to multiple optical fibers (4). The positioning bosses (11) have fiber-passing holes in the middle, so that each optical fiber (4) passes through the fiber-passing holes of a set of positioning bosses (11). The encapsulation layer (3) is used to cover the fingertip base (1) and the fiber array.

2. The integrated fiber Bragg grating fingertip tactile sensor according to claim 1, characterized in that, The encapsulation layer (3) is formed by mixing polydimethylsiloxane liquid and curing agent at a mass ratio of 10:1 and then curing them in a molding mold (5).

3. The integrated fiber Bragg grating fingertip tactile sensor according to claim 1, characterized in that, The surface of the fingertip base (1) is also fixedly provided with a plurality of fixed bosses (12), which are used to form a mechanical fit with the encapsulation layer (3).

4. The integrated fiber Bragg grating fingertip tactile sensor according to claim 1, characterized in that, Each optical fiber (4) is provided with multiple fiber Bragg grating sensing units (41). Each of the fiber Bragg grating sensing units (41) has a corresponding center reflection wavelength, and the center wavelengths of different fiber Bragg grating sensing units (41) on each fiber (4) are staggered from each other.

5. The integrated fiber Bragg grating fingertip tactile sensor according to claim 4, characterized in that, The fiber array includes two optical fibers (4). The surface of the fingertip base (1) is fixedly provided with two sets of corresponding positioning bosses (11). The two optical fibers (4) are arranged in parallel through the two sets of positioning bosses (11). Each optical fiber (4) is provided with four fiber Bragg grating sensing units (41).

6. A method for packaging and manufacturing an integrated fiber Bragg grating fingertip tactile sensor as described in any one of claims 1-5, characterized in that, Includes the following steps: S1. Complete the manufacturing of fingertip base (1), fingertip connector (2) and multiple optical fibers (4) respectively. The structure and size of the fingertip connector (2) are matched with the corresponding machine finger structure. The optical fiber (4) contains a fiber Bragg grating sensing unit (41) that meets the requirements. S2. Fix the fingertip base (1) to the fingertip connector (2) and pass multiple optical fibers (4) through the fiber holes on multiple sets of positioning bosses (11) so that each optical fiber (4) reaches the preset position. S3. The overall structure after step S2 is sent to the preset molding mold (5) and assembled. S4. Place the assembled molding mold (5) into the preset packaging alignment device (6), which is used to fix the molding mold (5). S5. Prepare a mixture for forming the encapsulation layer (3); S6. Inject the mixture into the cavity of the molding mold (5); S7. The mixture is injected into the completed molding mold (5) and kept in a stable environment for curing to obtain the encapsulation layer (3). S8. After curing, release the sealing alignment device (6) from fixing the molding mold (5) and demold to complete the manufacturing of the integrated fiber Bragg grating fingertip touch sensor.

7. The packaging and manufacturing method of the integrated fiber Bragg grating fingertip tactile sensor according to claim 6, characterized in that, In step S1, the fingertip base (1) and the fingertip connector (2) are manufactured by 3D printing; After the optical fiber (4) is prepared, it is connected to a fiber optic grating demodulator to detect whether the initial reflection peak of the fiber Bragg grating sensing unit (41) of each optical fiber (4) meets the requirements.

8. The packaging and manufacturing method of the integrated fiber Bragg grating fingertip tactile sensor according to claim 6, characterized in that, In step S1, when the fingertip base (1) is manufactured, each set of positioning bosses (11) of the fingertip base (1) is positioned such that one positioning boss (11) is located at one edge of the fingertip base (1); In step S2, when each optical fiber (4) reaches the preset position, both ends of it extend to the area outside the fingertip base (1), and one end is used as the lead-out end, which is provided with a connector structure for connecting to the fiber grating demodulator, and the other end is used as the reserved end and passes through the positioning boss (11) located at the edge of the fingertip base (1). In step S4, the encapsulation alignment device (6) also clamps both ends of each optical fiber (4) so ​​that each optical fiber (4) is in a pre-tensioned state. In step S8, after demolding, the reserved end of the optical fiber (4) is bonded and fixed to the corresponding positioning boss (11), and the optical fiber (4) located outside the positioning boss (11) is cut off to ensure that the cutting surface is flat, thus completing the manufacturing of the integrated fiber Bragg grating fingertip touch sensor.

9. The packaging and manufacturing method of the integrated fiber Bragg grating fingertip tactile sensor according to claim 6, characterized in that, In step S5, specifically: Add the polydimethylsiloxane liquid and the curing agent to a clean container at a mass ratio of 10:1, stir thoroughly to ensure uniform dispersion of the curing agent, and obtain a mixture. After stirring, place the mixture in a vacuum environment for 15 to 20 minutes to degas, allowing the bubbles in the mixture to rise and burst.

10. The packaging and manufacturing method of the integrated fiber Bragg grating fingertip tactile sensor according to claim 6, characterized in that, Following step S8, the following step is also included: S9. Connect the fiber (4) end of the integrated fiber Bragg grating fingertip touch sensor to the fiber grating demodulator and check whether the reflection peak of each fiber (4) meets the requirements when it is stationary; and check whether the reflection peak of each fiber (4) meets the requirements when it is working by pressing the surface of the encapsulation layer (3) on the corresponding side of the fiber (4).

Images