Flexible fiber bragg grating multi-dimensional tactile sensing method and system

Abstract

The application belongs to the technical field of optical fiber sensing and tactile perception, and discloses a flexible fiber grating multi-dimensional tactile perception method and system. The method encapsulates a seven-core optical fiber in a silicone elastomer, writes Bragg gratings in each fiber core, and locates the Bragg gratings at the same cross section to realize common point measurement. Temperature self-compensation is realized through multi-core wavelength drift common mode extraction. Normal force and biaxial tangential force are decoupled based on an elasticity mechanics transfer matrix and a weighted pseudo-inverse algorithm. Contact position is estimated by using peripheral fiber core differential response, and structured tactile data is output. The application does not need an additional reference grating, has a compact structure, has high force decoupling precision, and can provide reliable multi-dimensional tactile perception for robot precise operation.

Description

Technical Field

[0001] This invention belongs to the field of fiber optic sensing and tactile perception technology, specifically relating to a flexible multidimensional tactile perception method and system based on multi-core fiber Bragg gratings, which can be applied to scenarios such as robot precision grasping and human-machine safe interaction. Background Technology

[0002] Tactile sensing technology is a key support for robots to achieve safe interaction and precise operation, playing an irreplaceable role in fields such as intelligent manufacturing, medical rehabilitation, and human-robot collaboration. As robots develop towards greater precision and intelligence, simply acquiring a single normal contact force is no longer sufficient to meet the needs of complex operational tasks. Robot end effectors urgently need to simultaneously sense normal pressure and tangential shear force in order to achieve a comprehensive judgment of grasping force, sliding tendency, and contact posture.

[0003] While existing piezoresistive and capacitive tactile sensors have achieved a certain degree of force measurement, they have inherent limitations. Piezoresistive sensors are extremely sensitive to changes in environmental electromagnetic fields and humidity, resulting in significant signal drift in industrial applications and making it difficult to guarantee stability over long periods. Although capacitive sensors offer higher sensitivity, their complex multi-layer electrode structure limits the miniaturization of the sensing unit, and temperature fluctuations significantly affect the dielectric constant, leading to measurement errors.

[0004] Fiber Bragg grating sensors, with their advantages of electromagnetic interference resistance, electrical insulation, small size, and wavelength division multiplexing capability, show promising application prospects in the field of tactile sensing. Chinese patent CN120869405A discloses a biomimetic fingertip tactile sensor based on fiber Bragg gratings and its sensitivity-enhancing packaging method. This scheme uses a single-core fiber Bragg grating encapsulated in a three-layer biomimetic composite structure, enhances stress transmission efficiency through a micropillar array structure within a flexible contact layer, and eliminates temperature drift using a reference grating differential algorithm. This scheme has achieved some progress in normal pressure detection sensitivity, with its micropillar array increasing stress transmission efficiency to more than three times that of traditional designs. However, the technical solution in prior art 1 can only detect normal pressure in a single dimension and cannot sense the tangential shear force component. When a robot performs a precise grasping task requiring judgment of sliding trends, single-dimensional force information is insufficient to support safe and reliable grasping control decisions. Furthermore, this scheme uses an independent reference grating for temperature compensation, requiring additional optical path space and calibration processes, which is difficult to implement in miniaturized packaging.

[0005] From a broader perspective of existing technologies, while multidimensional force sensing schemes based on multiple independent single-core optical fibers can distinguish the direction of force by arranging fibers in different orientations, the extraction and packaging of multiple fibers significantly increases structural complexity and manufacturing costs. Furthermore, ensuring spatial coherence among the fibers is difficult, leading to inconsistencies in measurement points and introducing systematic errors. Although intensity modulation-based fiber optic tactile sensing schemes have a simple structure, the intensity signal is susceptible to fluctuations in light source and connection losses, resulting in lower long-term stability compared to wavelength-encoded schemes.

[0006] In recent years, the development of multi-core fiber technology has brought new opportunities to the field of fiber optic sensing. Multi-core fibers integrate multiple spatially separated cores within a single fiber, each capable of independently transmitting optical signals and having its own Bragg grating. Theoretically, this allows for the acquisition of multi-dimensional strain information while maintaining structural compactness. However, systematic technical solutions are still lacking for how to effectively utilize the composite response characteristics of multi-core fiber gratings in flexible elastomers to achieve precise decoupling of three-dimensional forces, particularly how to establish a reliable elastic mechanical transfer model to map multi-channel wavelength drift information into independent force components, and how to fully utilize the spatial co-location characteristics of multi-core fibers to achieve temperature self-compensation without additional reference elements.

[0007] In summary, existing fiber Bragg grating tactile sensing technology faces the following pressing technical challenges: First, its ability to simultaneously perceive multi-dimensional forces is insufficient. Most existing solutions can only detect normal forces and cannot simultaneously sense tangential shear force components, making it difficult to meet the requirements of precise robotic operations for judging grasping force and sliding trends. Second, the accuracy of force decoupling is limited by sensor configuration and decoupling algorithms. The spatial coherence of multiple independent fiber optic solutions is difficult to guarantee, and the introduced geometric errors directly affect the accuracy of decoupling. Third, temperature compensation schemes rely on additional reference elements, increasing structural complexity and miniaturization difficulty. How to achieve high-precision independent calculation of normal and biaxial tangential forces and provide rich tactile information output while maintaining the miniaturization and structural simplicity of flexible tactile units is a core technical challenge that needs to be overcome in this field. Summary of the Invention

[0008] To address the technical challenges of existing fiber Bragg grating tactile sensors, such as the inability to simultaneously sense normal pressure and tangential shear force, low multidimensional force decoupling accuracy, and the increased miniaturization difficulty caused by temperature compensation schemes, this invention provides a flexible fiber Bragg grating multidimensional tactile sensing method, comprising the following steps: Multi-core fiber optic grating signal acquisition steps: Seven-core optical fibers are encapsulated in a silicone rubber elastomer to form a flexible tactile unit. The seven-core optical fiber includes a central core and six peripheral cores evenly distributed along the circumference of the cladding. Each core is engraved with a Bragg grating, and each grating region is located at the same cross-sectional position. When an external force is applied, the deformation of the elastomer is transmitted to the seven-core optical fiber to generate axial strain and bending strain. The Bragg wavelength shift of the seven grating channels is synchronously acquired through a wavelength division multiplexing demodulation module.

[0009] Temperature strain common-mode separation steps: common-mode components are extracted from the wavelength drift of the seven grating channels. The weighted average of the seven wavelength drifts is taken as the temperature-induced common-mode drift component. The common-mode drift component is subtracted from the original wavelength drift of each channel to obtain the pure strain wavelength drift vector.

[0010] Steps for constructing the elasticity transfer matrix: Based on the elasticity model of silicone rubber elastomer, establish the transfer matrix between the pure strain wavelength drift vector and the three-dimensional force components. Each element in the transfer matrix is ​​determined by the geometric parameters of the elastomer, the elastic modulus, the position coordinates of each fiber core in the cross section of the optical fiber, and the strain sensitivity coefficient of the fiber grating.

[0011] Multidimensional force decoupling solution steps: Perform weighted pseudo-inverse operation on the transfer matrix to map the pure strain wavelength drift vector into a three-dimensional force vector containing normal force components and biaxial tangential force components, thereby realizing the independent solution of normal force and tangential force.

[0012] Tactile data fusion and output steps: Calculate the force vector magnitude and direction angle based on the obtained three-dimensional force vector, estimate the contact position by combining the differential response modes of the six peripheral fiber core gratings, and encapsulate the three-dimensional force component time-series curves, contact position coordinates and force vector visualization data into a structured tactile data package for output.

[0013] The present invention also provides a flexible fiber optic grating multidimensional tactile sensing system, including a multi-core fiber optic grating signal acquisition module, a temperature strain common-mode separation module, an elastic mechanical transfer matrix construction module, a multidimensional force decoupling solution module, and a tactile data fusion and output module. Each module corresponds to the corresponding step of the above method, and together realizes the three-dimensional force sensing and structured tactile data output of the flexible tactile unit.

[0014] The beneficial effects of this invention are as follows: First, by using a seven-core optical fiber as the sensing carrier, the spatial distribution of the seven cores within a single fiber at the same cross-sectional position enables the common-point measurement of multi-dimensional force information. Compared to the multi-independent fiber scheme, this significantly reduces structural complexity and eliminates systematic errors caused by inconsistent measurement points. The extraction and packaging process of a single fiber is far simpler than that of multiple fibers, which is beneficial for the miniaturization and mass production of flexible tactile units. Second, temperature self-compensation is achieved by utilizing the common-mode component of wavelength drift from a multi-core grating. Since the spatial scale of the seven cores within the same fiber cross-section is much smaller than the scale of temperature field changes, the temperature-induced wavelength drift is highly consistent across channels. Temperature cross-sensitivity can be effectively eliminated through common-mode extraction and channel-by-channel subtraction, eliminating the need for additional reference gratings and independent optical paths, thus reducing the difficulty of miniaturization and system cost. Third, normal force and biaxial tangential force are realized based on the elasticity transfer matrix and weighted pseudo-inverse algorithm. The independent solution of the transfer matrix fully considers the directional response of bending strain to the spatial position difference of each fiber core in the cross-section of the optical fiber. The weighted pseudo-inverse method adaptively adjusts the weights according to the signal-to-noise ratio of each channel to improve the robustness of the solution. The truncated singular value decomposition regularization further ensures the numerical stability, and the force decoupling accuracy is significantly improved compared with the traditional method. Fourth, the contact position is estimated by the differential response mode of the outer fiber core. The wavelength drift difference of the symmetrical fiber core pair reflects the degree of bending of the elastomer in different directions, and then the eccentric position of the force application point on the sensor surface is calculated, providing a spatial information dimension that traditional force sensors do not have for robot grasping control. Fifth, the system is designed with a residual feedback closed-loop correction mechanism. The solution residual of the force decoupling module can drive the online update of the transfer matrix in reverse, so that the system can adapt to long-term drift factors such as creep of elastomer materials and environmental changes, and maintain the stability of measurement accuracy during long-term operation. Attached Figure Description

[0015] Figure 1 This is a flowchart of the flexible fiber optic grating multidimensional tactile sensing method of the present invention.

[0016] Figure 2 This is an architectural diagram of the flexible fiber optic grating multidimensional tactile sensing system of the present invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be noted that the following embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Equivalent modifications and improvements made by those skilled in the art based on the technical solutions of the present invention should be included within the scope of protection of the present invention.

[0018] like Figure 1As shown, this invention provides a multi-dimensional tactile sensing method for flexible fiber Bragg gratings. This method achieves high-precision independent calculation of normal and biaxial tangential forces by the flexible tactile unit through deep coupling and collaboration of five core steps: multi-core fiber Bragg grating signal acquisition, temperature and strain common-mode separation, elasticity transfer matrix construction, multi-dimensional force decoupling solution, and tactile data fusion and output. In one embodiment of this invention, the data processing link of the overall method forms a closed-loop architecture, and the calculation residuals of subsequent steps can be used to adjust the parameter configuration of preceding steps, thereby continuously optimizing the tactile sensing accuracy during long-term operation.

[0019] Step S1: Multi-core fiber Bragg grating signal acquisition. The core objective of this step is to encapsulate the seven-core fiber within a silicone rubber elastomer to form a flexible tactile unit, and simultaneously acquire the Bragg wavelength shift of the seven grating channels through a wavelength division multiplexing demodulation module. Preferably, in this embodiment, the seven-core fiber is a standard communication-grade germanium-boron co-doped seven-core fiber, with one central core located at the geometric center of the fiber, and six peripheral cores arranged at equal angular intervals along the circumference of the cladding with the central core as the center. The included angle between adjacent peripheral cores is 60 degrees. The radial spacing between each peripheral core and the central core is 42 μm, the fiber cladding diameter is 125 μm, and the coating diameter is 250 μm. The mode field diameter of each core is approximately 6.5 μm, and the numerical aperture is approximately 0.14. These parameters are compatible with standard single-mode fiber, facilitating interfacing with existing fiber Bragg grating demodulation systems. In one embodiment of the present invention, each fiber core is etched with a femtosecond laser point-by-point writing process using Bragg gratings. This process has the advantage over the traditional ultraviolet laser phase mask method in that it allows for independent writing of each fiber core in a multi-core fiber, and the writing process does not require stripping the coating layer, which helps maintain the mechanical strength of the fiber. The grating area length is 8 mm, and the center positions of each grating area are precisely aligned along the fiber axis to ensure they are located in the same cross-section. The writing alignment accuracy is better than 10 μm, thereby achieving co-point measurement in a mechanical sense. To achieve wavelength division multiplexing demodulation, the center wavelengths of the seven gratings are allocated at equal intervals. In this embodiment, the center wavelength of the central fiber core grating is set to 1545 nm, and the center wavelengths of the six peripheral fiber core gratings increase from 1548 nm to 1563 nm in 3 nm increments. The wavelength spacing between the gratings is sufficient to avoid crosstalk.

[0020] In the fabrication of the flexible haptic unit, the silicone rubber elastomer is selected from polydimethylsiloxane material with a Shore hardness of A30. Its elastic modulus is approximately 1.2 MPa, Poisson's ratio is approximately 0.48, and elongation at break is greater than 400%, allowing it to withstand repeated bending and stretching without permanent deformation. The elastomer has dimensions of 12mm × 12mm × 6mm and a total weight of approximately 1.2g, suitable for miniaturized installation spaces such as robot fingertips. The selection of silicone rubber material has a significant impact on system performance: too low a Shore hardness will cause excessive deformation of the elastomer, causing the optical fiber to bend beyond the linear response range; too high a Shore hardness will reduce force transmission efficiency and decrease sensitivity. In this embodiment, the A30 hardness achieves a good balance between sensitivity and linear range, with a linear range of 0 to 5N for normal force and 0 to 3N for tangential force. The seven-core optical fiber passes centrally along the length of the elastomer, with the encapsulation depth located at 50% of the elastomer thickness, i.e., 3mm from the top surface. The encapsulation process employs a two-step casting and curing method: First, a layer of silicone rubber is cast into the bottom of a customized PTFE mold and cured at room temperature to a semi-cured state. At this stage, the silicone rubber surface still retains adhesiveness, allowing for a good interface bond with subsequent casting layers. Then, the seven-core optical fiber is accurately positioned at a preset depth, and a pre-tension force of approximately 50 mN is applied using a fiber optic clamp to eliminate fiber slack and ensure synchronous deformation of the fiber and elastomer. Finally, a second layer of silicone rubber is cast to cover the fiber and cured at 60°C for 2 hours to complete the overall encapsulation. The two ends of the optical fiber extend from the side of the elastomer and are connected to standard single-core optical fiber patch cords via fan-in / fan-out devices.

[0021] The wavelength division multiplexing (WDM) demodulation module employs a scanning fiber Bragg grating demodulator with a scanning rate of 2 kHz, a wavelength range covering 1525 nm to 1570 nm, and a wavelength resolution of 0.5 μm, capable of simultaneously resolving seven wavelength channels. When an external force is applied to the upper surface of the flexible tactile unit, the elastic body deforms and transmits the strain to the seven internal fiber cores. The central fiber core is primarily subjected to axial compressive strain, while the outer fiber cores, depending on their position in the cross-section, are simultaneously affected by the superposition of axial and bending strains. The Bragg wavelength of each grating drifts with changes in strain state, and the demodulator measures and records the wavelength drift of the seven channels in real time. Preferably, during the signal preprocessing stage, the original wavelength drift is subjected to a fourth-order Butterworth bandpass filter with a bandwidth of 0.5Hz to 50Hz. A low-frequency cutoff frequency of 0.5Hz is used to remove baseline drift caused by slow ambient temperature drift, and a high-frequency cutoff frequency of 50Hz is used to suppress electronic noise and mechanical vibration noise from the demodulator. The filtered seven-channel wavelength drift data constitute the original wavelength drift vector. This serves as the input for the subsequent temperature strain common-mode separation step.

[0022] The seven-channel wavelength drift vectors output in this step carry both temperature and strain effects. Since the spatial scale of the seven fiber cores within the same cross-section is much smaller than the spatial scale of the temperature field, the temperature changes experienced by each fiber core can be considered identical. This co-location characteristic provides the physical basis for temperature self-compensation based on common-mode extraction in subsequent steps. It is important to note that the core advantage of seven-core fiber compared to multiple independent single-core fibers lies in its spatial co-location. The spatial deviation of the seven measurement points at the same cross-section along the fiber axis is determined by the fiber manufacturing precision, typically not exceeding 1µm, far superior to the precision level achievable by manual positioning of multiple independent fibers. This micrometer-level spatial co-location not only ensures a high degree of consistency in the temperature response of each channel but also guarantees the consistency of the geometric reference for strain analysis of each fiber core in the elasticity model. Furthermore, the performance of the fan-in and fan-out devices in this step is crucial to the overall system performance. Preferably, the fan-in and fan-out devices employ end-face grinding alignment coupling technology, with insertion loss of each channel not exceeding 1.5dB and inter-channel crosstalk below -40dB, ensuring the independence of each channel signal and the signal-to-noise ratio during wavelength division multiplexing demodulation.

[0023] Step S2: Temperature-Strain Common-Mode Separation. The core objective of this step is to separate the temperature-induced common-mode drift component and the strain-induced differential-mode response component from the original wavelength drift vector of the seven channels, achieving temperature self-compensation. In one embodiment of the present invention, the physical basis for temperature compensation is that, since the seven fiber cores in a seven-core optical fiber are located within the same fiber cross-section and their spatial spacing is only on the order of micrometers, the temperature changes of each fiber core are essentially the same under the action of a macroscopic temperature field. Therefore, the temperature-induced wavelength drift manifests as a common-mode signal among the seven channels. In contrast, the strain effect caused by external force presents a differentiated differential-mode response mode due to the different positions of each fiber core in the cross-section. For example, the normal compressive force causes the axial strain of the central fiber core to be the largest, while the axial strain of the peripheral fiber cores exhibits directional differences due to the superposition of bending effects.

[0024] Specifically, common-mode drift components The calculation method is a weighted average of the wavelength shifts of the seven channels: ,in, For the first The original wavelength shift of each grating channel, in pm; For the first The normalized weight coefficients of each channel are dimensionless and satisfy the following conditions: ; Subscript The value range is from 1 to 7. Corresponding to the central fiber core, to Corresponding to the six outer fiber cores. Weighting coefficients. Based on the temperature sensitivity coefficient of each fiber core Normalization determination, i.e. ,in The unit is pm / ℃, determined through calibration experiments. In this embodiment, since the temperature sensitivity coefficients of the seven fiber cores differ very little (deviation not exceeding 3%), the weighting coefficients are approximately equal in weight. When there are significant differences in temperature sensitivity coefficients, using unequal weighting coefficients can further improve the accuracy of temperature compensation.

[0025] The pure strain wavelength shift is obtained by subtracting the common-mode shift component from the original wavelength shift of each channel. ,in, For the first The pure strain wavelength shift of each channel, expressed in pm. The pure strain wavelength shifts of the seven channels constitute the pure strain wavelength shift vector. .

[0026] Preferably, the present invention introduces a temperature compensation quality assessment mechanism during the common-mode separation process: the ratio of the mean to the standard deviation of each component in the pure strain wavelength drift vector is calculated as the compensation quality index. .definition for: ,in, Seven pure strain wavelength shifts The arithmetic mean of , in pm; This represents the standard deviation of the seven pure strain wavelength shifts, in pm. When... A value close to 1 indicates that the common-mode component has been sufficiently removed. When the temperature falls below a preset threshold of 0.85, the system issues a temperature compensation quality alarm, indicating potential interference from a non-uniform temperature field or sensor packaging defects. In this case, the weighting coefficient can be adaptively adjusted. Alternatively, an alternative differential temperature compensation strategy can be used to maintain compensation accuracy. In one embodiment of the present invention, the temperature compensation residual does not exceed 2% of the pure strain signal amplitude, and the measured temperature compensation accuracy is better than 0.3℃ within the temperature range of 0℃ to 50℃.

[0027] The pure strain wavelength drift vector output in this step Including only strain information caused by external forces effectively eliminates the temperature cross-sensitivity effect, providing reliable input data for the establishment of the elasticity transfer matrix and force decoupling calculations in subsequent steps. It is worth noting that the temperature self-compensation mechanism is entirely based on the spatial co-location characteristics of the multi-core optical fiber itself, without the need for additional reference gratings or independent temperature sensing elements. This characteristic allows the flexible haptic unit to maintain a compact structure, which is beneficial for miniaturized applications such as robot fingertips.

[0028] Preferably, in one embodiment of the present invention, after the common-mode separation step, adaptive noise suppression processing is further applied to the pure strain wavelength drift vector. Specifically, a covariance matrix is ​​constructed using the statistical correlation between the seven channel signals, and the main strain mode components are extracted based on principal component analysis. Higher-order components with an energy percentage less than 1% of the total energy are treated as noise and removed. This method can effectively suppress demodulator sampling noise and random disturbances caused by micro-vibrations of the elastic body, while retaining the useful signal components of the force response, providing higher-quality input data for the subsequent force decoupling step. Experimental results show that after principal component analysis denoising, the average signal-to-noise ratio of the seven-channel signals is increased from the original 32dB to approximately 45dB, and the accuracy of force decoupling is significantly improved.

[0029] Step S3: Construction of the elasticity transfer matrix. The core objective of this step is to establish a quantitative mapping relationship between the pure strain wavelength drift vector and the three-dimensional force components, i.e., the transfer matrix. In one embodiment of the present invention, the construction of the transfer matrix is ​​based on the continuum mechanics model of silicone rubber elastomer and the strain-wavelength response characteristics of fiber Bragg gratings.

[0030] First, define the coordinate system and force components: taking the center of the upper surface of the flexible tactile unit as the origin, define the coordinate system along two orthogonal directions on the surface of the elastic body. shaft and The axis is perpendicular to the surface and downwards. Axis. The three-dimensional force vector is denoted as... ,in and respectively along shaft and Tangential force component in the axial direction, For along Normal force components along the axial direction, all in N.

[0031] When normal force When applied to the upper surface of an elastomer, a compressive strain field is generated inside the elastomer. At the fiber encapsulation depth, the central fiber core experiences axial compressive strain. It can be calculated based on the theory of elasticity: ,in, The elastic modulus of the silicone rubber elastomer is taken as , in this embodiment . MPa; The equivalent bearing area is related to the surface area of ​​the elastomer and the fiber encapsulation depth. In this embodiment... The negative sign indicates that the direction of compressive strain is parallel to the direction of... The positive axis is opposite. The outer fiber core also experiences axial compressive strain under the normal force, due to its radial distance offset from the fiber center. In addition, there is a small additional strain caused by the transverse Poisson effect of the elastomer, but this additional amount usually does not exceed 5% of the principal strain.

[0032] When tangential force or When applied to the upper surface of an elastomer, a shear strain field is generated within the elastomer, causing the optical fiber to bend. This bending results in positive and negative strain components in the outer core, which deviates from the neutral axis, related to the bending direction. Let the first... The angular position of the outer core in the cross-section of the optical fiber is (by (With the positive axis direction being 0 degrees and counterclockwise being positive), the fiber core is subjected to tangential force. The bending strain components under the action are: ,in, The radial distance between the outer core and the central core is shown in this embodiment. ; For the first The angular position of the outer core fiber in the cross-section, measured in rad. The value range is 1 to 6, corresponding to ; Let be the Young's modulus of the optical fiber; for silica optical fiber, take . GPa; The moment of inertia of the fiber cross-section is given in this embodiment. Given an optical fiber cladding diameter of 125 μm, the calculation is as follows: Similarly, tangential force The resulting bending strain components are: The meanings of each parameter are the same as above. Reflects directional tangential force in the first The bending contribution projection of the peripheral fiber core is considered. Since the central fiber core is located at the geometric center of the fiber, i.e., the bending neutral axis, its bending strain component is theoretically zero under ideal pure bending conditions. Therefore, the central fiber core primarily responds to normal compressive forces and is insensitive to tangential forces. This characteristic makes the central fiber core a natural dedicated sensing channel for normal forces. The peripheral fiber core, however, is sensitive to the bending direction due to its deviation from the neutral axis. The sign and magnitude of its bending strain component depend on the relative relationship between the angular position of the fiber core in the cross-section and the bending direction. This functional complementarity between the central and peripheral fiber cores in terms of mechanical response constitutes the physical basis of the multidimensional force sensing of this invention.

[0033] The combined effect of normal and tangential forces, the first The total strain of the outer core is: The relationship between the wavelength drift and axial strain of a Bragg grating is as follows: ,in, For the first The Bragg center wavelength of the outer core grating, in nm; The effective elastic-optical coefficient of the optical fiber is given by the formula for germanium-doped silica optical fiber. Dimensionless. Substituting and rearranging the above formulas, the pure strain wavelength shift of the seven channels can be expressed as a linear combination of three-dimensional force components: ,in, It is a 7×3 dimensional transfer matrix, whose elements are derived from the elastic modulus of the elastic body. Equivalent bearing area Radial distance of each fiber core , angle and position grating center wavelength and effective elastic coefficient Determined jointly. Transfer matrix. The first column corresponds to the normal force = The sensitivity coefficient vector, the second and third columns respectively correspond to direction and The sensitivity coefficient vector of tangential force.

[0034] Preferably, in practical applications, the transfer matrix also needs to be calibrated through a calibration experiment. In one embodiment of the present invention, the calibration process uses a six-degree-of-freedom force sensor as a reference force source, along which the force is applied respectively. Known force loads were applied to the flexible haptic unit in three directions, ranging from 0 to 5 N with a step size of 0.1 N. The wavelength drift response of the seven grating channels under each force load was recorded, and the calibration transfer matrix was obtained by least squares fitting. The theoretical transfer matrix and the calibration transfer matrix are then combined using a weighted fusion method to correct for simplification errors in the elasticity model. The fusion weights can be determined using cross-validation. In this embodiment, the fusion ratio of the theoretical matrix to the calibration matrix is ​​3:7, indicating a bias towards experimental calibration data.

[0035] Furthermore, this invention considers the impact of the nonlinear effects of elastomer materials on the accuracy of the transfer matrix. Under large deformation conditions, silicone rubber elastomers may deviate from the linear elasticity assumption, causing slight changes in the elements of the transfer matrix with the magnitude of the force load. Therefore, in one embodiment of this invention, a piecewise linearization strategy is employed, dividing the force range into several sub-intervals and calibrating a transfer matrix within each sub-interval. In this embodiment, the range of 0 to 5N is divided into three sub-intervals: 0 to 1N, 1 to 3N, and 3 to 5N. The transfer matrix of the corresponding sub-interval is automatically selected for force decoupling calculation based on the estimated value of the current force load. A weighted smooth transition is used when switching between intervals to avoid abrupt discontinuities in the force output, with a transition bandwidth of ±10% of the boundary between adjacent intervals. This piecewise calibration strategy effectively compensates for the nonlinear effects of the material without significantly increasing the calibration workload, ensuring that the transfer matrix maintains high descriptive accuracy across the entire range.

[0036] In one embodiment of the present invention, the condition number of the transfer matrix is ​​also an important indicator for evaluating the force decoupling performance. The condition number reflects the degree to which the transfer matrix amplifies the input error during the numerical inversion process; a smaller condition number indicates better numerical stability. In this embodiment, the condition number of the transfer matrix is ​​approximately 8.5, which is at a relatively optimal level. This is mainly due to the symmetrical spatial arrangement of the outer cores in the seven-core optical fiber, which provides sufficient spatial resolution for the identification of normal force and biaxial tangential force.

[0037] Step S4: Multidimensional Force Decoupling Solution. The core objective of this step is to inversely map the pure strain wavelength drift vector into a three-dimensional force vector based on the transfer matrix, thereby achieving independent calculation of the normal force and the biaxial tangential force. Due to the transfer matrix... For a 7×3 dimensional overdetermined matrix (with more equations than unknowns), direct inversion is not feasible. Therefore, pseudo-inversion operation is required to perform the optimal solution in the least squares sense.

[0038] In one embodiment of the present invention, to improve the robustness and accuracy of force decoupling, a weighted pseudo-inverse method is used instead of the ordinary pseudo-inverse. First, a 7×7 dimensional diagonal weighted matrix is ​​constructed. ,in, For the first The weight values ​​for each channel are determined based on the signal-to-noise ratio of that channel. Sure, Dimensionless; Defined as the ratio of the root mean square (RMS) of the pure strain wavelength drift of the channel to the RMS of the noise, it is dimensionless. Channels with higher signal-to-noise ratios (SNRs) receive greater weight in force decoupling, thereby reducing the impact of noise from low SNR channels on the solution results. In one embodiment of the present invention, It is calculated by collecting background noise data under no-load conditions and signal data when a standard force load is applied.

[0039] The weighted transfer matrix and its pseudoinverse are: , The formula for solving the three-dimensional force vector is: ,in, The calculated three-dimensional force vector, in units of N; It is a 3×7 dimensional weighted pseudo-inverse matrix.

[0040] Furthermore, to suppress the noise amplification effect when the condition number of the transfer matrix is ​​large in certain directions, this invention introduces a truncated singular value decomposition regularization method in the pseudo-inverse operation. Specifically, for the weighted transfer matrix... Perform singular value decomposition: ,in, It is a 7×7 dimensional orthogonal matrix. It is a 7×3 dimensional diagonal matrix, whose diagonal elements It is a singular value. It is a 3×3 orthogonal matrix. The truncation threshold is set to 1% of the maximum singular value, that is, when... The condition number of the propagation matrix is ​​set to zero to construct a regularized pseudo-inverse matrix. In this embodiment, the condition number of the propagation matrix is... The value is approximately 8.5, indicating that the matrix has good numerical stability. Truncation regularization is mainly used as a safety mechanism to prevent the solution from diverging under extreme conditions.

[0041] Preferably, the force decoupling solution step further includes a residual feedback mechanism: the calculated three-dimensional force vector is fed back. Substitute into the transfer matrix to calculate the predicted wavelength drift vector and compared with the actual measured pure strain wavelength drift vector Compare and calculate the residual vector. and root mean square value of residuals .when When the threshold is exceeded (2pm in this embodiment), the system automatically triggers the online correction process of the transfer matrix, updating the elements in the transfer matrix using the recursive least squares method to adapt to long-term drift effects such as creep and aging of the elastomer material. This residual feedback mechanism constitutes the closed-loop correction loop in the method, enabling the system to maintain the stability of force decoupling accuracy during long-term operation.

[0042] In one embodiment of the present invention, the forgetting factor of the recursive least squares method is set to 0.998. This parameter controls the decay rate of historical data on the current matrix update. The closer the forgetting factor is to 1, the stronger the system's memory of historical data, and the smoother the matrix update, but the weaker the tracking ability; the smaller the forgetting factor, the faster the system responds to the latest data, but the update process is more susceptible to noise interference. In this embodiment, 0.998 is selected as an optimized result that balances tracking speed and stability, and can adapt to the slow changes in the creep rate of elastomer materials, which is usually on the order of 0.1% / h. Preferably, the forgetting factor can be adaptively adjusted according to the residual change trend: when the residual continues to rise, the forgetting factor is reduced to accelerate tracking; when the residual stabilizes, the default value is restored to maintain stability.

[0043] Furthermore, this invention introduces a force component rationality verification mechanism during the multidimensional force decoupling solution process. In typical applications of robot tactile perception, the normal force is usually much larger than the tangential force, with the ratio of normal to tangential force typically ranging from 2:1 to 10:1. When the calculated force component ratio deviates from this range, the rationality verification mechanism issues an alarm indicating possible abnormal contact or sensor malfunction, and can selectively constrain and correct the force components of abnormal frames to prevent obviously unreasonable force output values ​​from being transmitted to the robot control system, leading to malfunctions.

[0044] Step S5: Tactile data fusion and output. The core objective of this step is to further process the calculated three-dimensional force vector into structured tactile data that can be directly used by the robot control system. This includes three sub-steps: force vector feature calculation, contact position estimation, and data encapsulation and output.

[0045] Force vector magnitude The formula for calculating the direction angle is: ,in, The magnitude of the three-dimensional force vector, in N; These represent the force components along the three coordinate axes, in N. The zenith angle of the force vector. and azimuth Defined as: ,in, Force vector and The angle between the positive axes and the positive axes has a range of values. The unit is rad. When the force is close to 0, it indicates that the force direction is approximately perpendicular to the sensor surface (pure normal force). near The force direction is approximately parallel to the sensor surface (pure tangential force). For force vectors in Plane projection and The angle between the positive axes and the positive axes has a range of values. The unit is rad, which reflects the direction of the tangential force; It is the arctangent function in the four quadrants.

[0046] Contact position estimation is an important functional extension of this invention. In one embodiment, the position of the contact force application point on the surface of the elastomer is estimated using a spatially symmetrical arrangement of six peripheral fiber cores. The physical principle is as follows: when the force application point deviates from the center of the elastomer surface, the strain distribution inside the elastomer is no longer symmetrical about the center. The fiber core closer to the force application point experiences greater strain than the fiber core farther away. This asymmetrical strain distribution is reflected in the wavelength drift difference between the peripheral fiber core pairs. The six peripheral fiber cores are divided into three symmetrical fiber core pairs according to their spatial positions: the first group contains two fiber cores with angular positions of 0° and 180°, the second group contains two fiber cores with angular positions of 60° and 240°, and the third group contains two fiber cores with angular positions of 120° and 300°. The wavelength drift difference between each symmetrical fiber core pair is calculated. ,in, For the first The difference between symmetrical fiber core pairs, The value range is from 1 to 3; and The first The fiber core numbers within the group whose angular position is in the first and second halves of the circumference. Difference. This reflects the degree of bending of the elastomer in this direction, and the degree of bending is proportional to the eccentricity of the force application point relative to the center. By vector synthesis of three sets of differences, the eccentricity of the contact force application point relative to the center of the elastomer surface can be estimated. and eccentricity direction angle : , , in, The eccentricity calibration coefficient, in mm / pm, is determined through a calibration experiment by applying a force load at a known eccentric position. In this embodiment... mm / pm; This represents the eccentricity of the contact point, in mm. The eccentricity direction angle is expressed in rad. The Cartesian coordinates of the contact position are... In this embodiment, the accuracy of contact position estimation is better than 0.5 mm in the central region of the elastomer surface (within a radius of 3 mm).

[0047] Regarding data encapsulation and output, this step organizes the above calculation results into a structured haptic data package, which includes the following fields: three-dimensional force component time-series data. Force vector magnitude and direction angle time series data Contact position coordinates time series data and temperature compensation quality indicators Sum of residual indices The tactile data packets are output in a standardized format (JSON format is used in this embodiment). The data update frequency is consistent with the demodulator sampling frequency of 2kHz. They can be transmitted to the robot's host computer control system via TCP / IP or ROS message interface for use by high-level control strategies such as gripping force control, sliding detection, and contact posture judgment.

[0048] Preferably, this step also includes a force vector visualization function: the magnitude and direction of the force vector are displayed in real time as a three-dimensional arrow graphic on the human-machine interface, the distribution of the contact position on the surface of the elastic body is displayed in the form of a heat map, and the dynamic change history of each force component and the force vector magnitude is displayed in the form of a time-series curve. The visualization interface refreshes at 30 frames per second, providing operators with intuitive tactile feedback.

[0049] Furthermore, this invention introduces a tactile event detection mechanism in the data encapsulation output. This mechanism automatically identifies typical tactile event types based on the temporal and amplitude characteristics of the three-dimensional force signal, including contact establishment events (force vector magnitude rises from zero and exceeds the contact threshold, which is set to 0.05N in this embodiment), steady-state contact events (force vector magnitude remains within a stable range for more than 0.5s), sliding trend events (the ratio of tangential force to normal force exceeds 80% of the estimated static friction coefficient), and contact release events (force vector magnitude falls back below the contact threshold). Each tactile event is labeled with its occurrence time, duration, peak force, and event type, and appended to the structured tactile data packet. This tactile event information can be directly used in the decision-making logic of the robot's upper-level control strategy. For example, when a sliding trend event is detected, the normal gripping force can be increased in a timely manner to prevent the object from slipping, or when a contact release event is detected, the object can be determined to have been released and a re-grip process can be triggered.

[0050] In one embodiment of the present invention, the tactile data packet further includes system self-diagnostic information, specifically including parameters such as the signal strength and signal-to-noise ratio of each grating channel, temperature compensation quality index, force decoupling residual index, and condition number of the transfer matrix. This self-diagnostic information enables the robot control system to determine the reliability of the current tactile data and automatically switch to a conservative control strategy to ensure operational safety when the performance of the tactile sensor deteriorates. Each field in the data packet carries a unified timestamp, with a time synchronization accuracy better than 0.1ms, ensuring precise time alignment with the robot's motion control data.

[0051] like Figure 2As shown, the present invention also provides a flexible fiber optic grating multi-dimensional tactile sensing system, which includes a multi-core fiber optic grating signal acquisition module 1, a temperature strain common-mode separation module 2, an elasticity transfer matrix construction module 3, a multi-dimensional force decoupling solution module 4, and a tactile data fusion and output module 5. Each module corresponds one-to-one with the five steps in the above method embodiment.

[0052] The multi-core fiber Bragg grating signal acquisition module 1 comprises two sub-components: a flexible tactile unit and a wavelength division multiplexing (WDM) demodulation unit. The flexible tactile unit, as described in the method embodiment, includes a seven-core optical fiber encapsulated within a silicone rubber elastomer. The elastomer has external dimensions of 12mm × 12mm × 6mm, and the Bragg grating regions etched on each core of the seven-core optical fiber are located at the same cross-sectional position. The WDM demodulation unit uses a fiber Bragg grating demodulator with a scan rate of not less than 1kHz, equipped with fiber fan-in / fan-out devices to lead the signals from each core of the seven-core optical fiber to a standard single-core optical fiber channel for parallel demodulation. The output of this module is a digitized time-series data stream of the seven-channel Bragg wavelength shift. Preferably, the demodulation unit incorporates a 0.5Hz to 50Hz digital bandpass filter for signal preprocessing.

[0053] The temperature strain common-mode separation module 2 receives a seven-channel wavelength drift data stream, extracts the temperature common-mode drift component according to the weighted average method described in step S2 of the method embodiment, performs channel-by-channel subtraction, and outputs a pure strain wavelength drift vector. This module also includes a temperature compensation quality monitoring submodule, which calculates the compensation quality index in real time. It triggers an alarm when the indicator falls below a threshold. In one embodiment of the present invention, the module can be integrated into the embedded processor of a fiber Bragg grating demodulator to achieve low-latency processing, or it can be deployed on an external data processing computer for algorithm debugging and parameter adjustment.

[0054] Elasticity transfer matrix construction module 3 stores the transfer matrix that combines calculations based on elasticity theory with fitting from calibration experiments. This module comprises a parameter storage submodule and a matrix update submodule. The parameter storage submodule stores physical model parameters such as the elastic modulus, Poisson's ratio, geometric dimensions, position coordinates of each fiber core, and grating parameters of the elastomer. The matrix update submodule receives residual feedback signals from the multidimensional force decoupling solution module. When the solution residual exceeds a preset threshold, it updates the transfer matrix elements online using the recursive least squares method. Preferably, this module also stores multiple alternative transfer matrix schemes for different temperature ranges and different service life stages. It can automatically switch to the most suitable transfer matrix version based on the temperature estimate provided by the temperature strain common mode separation module and the cumulative system operation time, thereby adapting to the drift of transfer characteristics caused by material creep and changes in ambient temperature.

[0055] The multidimensional force decoupling solution module 4 receives the pure strain wavelength drift vector and transfer matrix, and performs three-dimensional force calculation according to the weighted pseudo-inverse method described in step S4 of the method embodiment. This module includes a built-in signal-to-noise ratio estimation submodule for real-time updating of the weight coefficients in the diagonal weighting matrix, and a truncated singular value decomposition regularization submodule to ensure numerical stability. The calculated output is a three-dimensional force vector. Simultaneously, the data is transmitted to the tactile data fusion and output module and the residual calculation submodule. The residual calculation submodule substitutes the calculated force vector back into the transfer matrix to calculate the predicted wavelength drift and compares it with the measured value. The residual information is fed back to the matrix update submodule of the elasticity transfer matrix construction module to form a closed-loop correction circuit. In one embodiment of the present invention, the single calculation delay of this module does not exceed 0.1ms, meeting the 2kHz real-time processing requirement. Preferably, the module also includes a force component rationality verification submodule and an abnormal frame marking submodule. The former judges the physical rationality of the force component based on the ratio range of the normal force to the tangential force, and the latter marks the data frames that fail the rationality verification as abnormal and adds an abnormal flag bit to the output data packet for the robot control system to refer to and judge.

[0056] The tactile data fusion and output module 5 receives the three-dimensional force vector, calculates the force vector magnitude, direction angle, and contact position coordinates according to the method described in step S5 of the method embodiment, and encapsulates all data into a structured tactile data packet, which is then output through the communication interface. This module includes a force vector calculation submodule, a contact position estimation submodule, a data encapsulation submodule, and a visualization submodule. The visualization submodule displays tactile information in real time on the human-computer interaction interface using various formats such as 3D arrows, heatmaps, and time-series curves. The data communication interface supports both TCP / IP and ROS message modes, and the data packet update frequency is synchronized with the demodulator sampling frequency.

[0057] The data flow relationship between the five modules is as follows: the output of the multi-core fiber optic grating signal acquisition module 1 serves as the input of the temperature strain common-mode separation module 2; the output of the temperature strain common-mode separation module 2, together with the transfer matrix provided by the elasticity transfer matrix construction module 3, serves as the input of the multi-dimensional force decoupling solution module 4; the output of the multi-dimensional force decoupling solution module 4 serves as the input of the tactile data fusion and output module 5; simultaneously, the residual signal of the multi-dimensional force decoupling solution module 5 is transmitted in reverse to the elasticity transfer matrix construction module 3 to achieve closed-loop correction. Data transmission between modules adopts a pipelined parallel processing architecture. After the previous module completes the processing of the current frame of data, it immediately transmits the result to the subsequent module and begins processing the next frame of data, thereby maximizing the real-time performance of the system.

[0058] Preferably, in one embodiment of the present invention, the system is further equipped with a parameter management module for unified management and storage of various parameters required for system operation, including fiber optic physical parameters (number of fiber cores, fiber core spacing, grating wavelength, grating length, etc.), elastomer material parameters (elastic modulus, Poisson's ratio, Shore hardness, geometric dimensions, etc.), signal processing parameters (filter cutoff frequency, sampling rate, noise reduction threshold, etc.), force decoupling parameters (weighting coefficients, regularization thresholds, residual triggering thresholds, etc.), and calibration data (calibration transfer matrix, eccentricity calibration coefficients, temperature sensitivity coefficients, etc.). The parameter management module supports parameter import / export and version management functions, and can easily migrate calibrated parameter sets to other haptic units of the same model, improving the deployment efficiency of multi-sensor arrays.

[0059] The system architecture of this invention follows the principles of modularity and loose coupling. Each functional module communicates through standardized data interfaces, facilitating subsequent functional expansion and upgrades. For example, to extend tactile perception from three-dimensional force to six-dimensional force perception including torque components, only the algorithm logic in the elasticity transfer matrix construction module and the multi-dimensional force decoupling solution module needs to be upgraded, while the signal acquisition module, temperature compensation module, and data output module can remain unchanged. As another example, to extend the flexible tactile unit from single-point perception to multi-point array perception, only the number of demodulation channels needs to be increased in the signal acquisition module and array synthesis logic added to the data fusion module; the core force decoupling algorithm framework can be directly reused.

[0060] To verify the performance of the flexible fiber Bragg grating multidimensional tactile sensing method and system of this invention, an experimental verification platform was built. The experimental platform includes the flexible tactile unit of this invention, a wavelength division multiplexing fiber Bragg grating demodulator (scanning rate 2kHz, wavelength resolution 0.5pm), a six-degree-of-freedom force sensor (range 0 to 10N, resolution 1mN) as a force reference, a three-axis precision displacement platform, and a data acquisition and processing computer.

[0061] In the normal force measurement experiment, a precision displacement platform is used along... A normal load of 0 to 5 N was applied to the upper surface of the flexible haptic unit along the axial direction, with a step size of 0.1 N. Experimental results show that the wavelength drift of the central fiber core grating exhibits a good linear relationship with the normal force, with a sensitivity of 18.5 pm / N and a linearity of [missing information]. The accuracy is 0.998. The decoupled normal force measurement accuracy is 0.8% of full scale, which means the resolution is approximately 0.04 N.

[0062] In the tangential force measurement experiment, along respectively... shaft and A tangential load of 0 to 3 N was applied to the flexible haptic unit along the axial direction. Experimental results show that the differential response of the outer fiber core grating is linearly related to the tangential force. The directional sensitivity is 6.2 pm / N. The directional sensitivity is 5.9 pm / N, and the crosstalk between the two directions is less than 3.5%. The measurement accuracy of tangential force is 1.2% of full scale.

[0063] In the three-dimensional force synchronous measurement experiment, a known combination of normal and tangential forces was applied simultaneously by a six-degree-of-freedom force sensor. The overall accuracy of the three-dimensional force decoupling algorithm was 1.5% of the full scale, which is about 40% higher than that of the ordinary least squares method without weighted pseudo-inverse.

[0064] In the temperature compensation performance verification experiment, the flexible haptic unit was placed in a temperature-controlled chamber, and the temperature was increased from 10℃ to 45℃ at a rate of 1℃ / min. The experimental results show that the temperature compensation residual of the multi-core common-mode temperature self-compensation method is equivalent to 0.25℃, which is far superior to the equivalent temperature compensation residual of 1.2℃ of the single-reference grating differential compensation scheme used in Comparative Document 1.

[0065] In the contact position estimation experiment, normal force loads were applied at 5×5 equally spaced marking points on the surface of the elastomer. The root mean square error of the contact position estimation results was 0.35 mm in the central 3 mm×3 mm area and 0.72 mm over the entire surface, which can meet the accuracy requirements of contact positioning in robot grasping control.

[0066] In the dynamic force response experiment, a sinusoidal normal force excitation with a frequency ranging from 0.1 Hz to 200 Hz was applied to the flexible tactile unit using a piezoelectric actuator, with a load amplitude of 1 N. Experimental results show that the system maintains a flat amplitude-frequency response characteristic within the frequency range of 0.1 Hz to 100 Hz, with an amplitude deviation not exceeding ±3% and a phase delay not exceeding 5°. In the 100 Hz to 200 Hz frequency band, the amplitude-frequency response exhibits approximately 12% attenuation due to the viscoelastic effect of the elastomer material, but it still meets the detection requirements for the vast majority of tactile events in robot operations.

[0067] In the long-term stability experiment, a constant normal force of 1N was continuously applied to the flexible haptic unit for 72 hours, and force measurements were recorded every hour. The results showed that without the online correction mechanism, the force measurement drift after 72 hours was 4.2%, mainly due to the creep effect of the silicone rubber elastomer. After enabling the recursive least squares online correction mechanism, the force measurement drift after 72 hours decreased to 0.8%, verifying the significant improvement effect of the residual feedback closed-loop correction mechanism on long-term stability.

[0068] In a comparative experiment with existing technologies, the multidimensional force decoupling accuracy of the present invention was compared with that of the single-dimensional normal force detection scheme in prior art document 1 and the three-dimensional force sensing scheme based on multiple independent optical fibers reported in the literature. Prior art document 1 can only measure normal force and lacks multidimensional force decoupling capability. The three-dimensional force sensing scheme based on 5 independent optical fibers reported in the literature has a normal force accuracy of 2.1%, a tangential force accuracy of 3.8%, and an inter-axis crosstalk of 5.2%. The present invention has a normal force accuracy of 0.8%, a tangential force accuracy of 1.2%, and an inter-axis crosstalk of 3.5%, which are significantly better than the multi-independent optical fiber scheme in all three indicators, while also significantly reducing structural complexity.

[0069] In summary, this invention achieves high-precision independent calculation of three-dimensional force information and structured output of rich tactile data by flexible tactile sensors through the deep coupling and synergy of technologies such as the common-point multidimensional sensing architecture of seven-core fiber gratings, the multi-core common-mode temperature self-compensation mechanism, the elastic mechanical transfer matrix and weighted pseudo-inverse force decoupling algorithm, and the differential contact position estimation of the peripheral fiber cores. This provides reliable multidimensional tactile sensing capabilities for precise robot operation and human-computer interaction.

[0070] The embodiments of the present invention are not limited to the specific embodiments described above. Those skilled in the art can make various equivalent changes or substitutions based on the technical solutions of the present invention, and all such changes or substitutions should be included within the protection scope of the present invention.

Claims

1. A flexible fiber optic grating multidimensional tactile sensing method, characterized in that, Includes the following steps: Multi-core fiber grating signal acquisition steps: Seven-core optical fibers are encapsulated in a silicone rubber elastomer to form a flexible tactile unit. The seven-core optical fiber includes a central fiber core and six peripheral fiber cores evenly distributed along the circumference of the cladding. Each fiber core is engraved with a Bragg grating, and each grating area is located at the same cross-sectional position. When an external force is applied to the flexible tactile unit, the deformation of the elastomer is transmitted to the seven-core optical fiber to generate axial strain and bending strain. The Bragg wavelength shift of the seven grating channels is synchronously acquired through a wavelength division multiplexing demodulation module. Temperature strain common-mode separation steps: common-mode components are extracted from the wavelength drift of the seven grating channels. The weighted average of the seven wavelength drifts is taken as the temperature-induced common-mode drift component. The common-mode drift component is subtracted from the original wavelength drift of each channel to obtain the pure strain wavelength drift vector. Steps for constructing the elasticity transfer matrix: Based on the elasticity model of silicone rubber elastomer, establish the transfer matrix between the pure strain wavelength drift vector and the three-dimensional force components. Each element in the transfer matrix is ​​determined by the geometric parameters of the elastomer, the elastic modulus, the position coordinates of each fiber core in the cross section of the optical fiber, and the strain sensitivity coefficient of the fiber grating. Multidimensional force decoupling solution steps: Perform weighted pseudo-inverse operation on the transfer matrix to map the pure strain wavelength drift vector into a three-dimensional force vector containing normal force components and biaxial tangential force components, thereby realizing the independent solution of normal force and tangential force; Tactile data fusion and output steps: Calculate the force vector magnitude and direction angle based on the obtained three-dimensional force vector, estimate the contact position by combining the differential response modes of the six peripheral fiber core gratings, and encapsulate the three-dimensional force component time-series curves, contact position coordinates and force vector visualization data into a structured tactile data package for output.

2. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, In the seven-core optical fiber, six peripheral cores are arranged at equal angular intervals along the circumference of the cladding with the central core as the center. The included angle between adjacent peripheral cores is 60 degrees. The radial spacing between each peripheral core and the central core is 35um to 50um. The grating length of each grating is 5mm to 10mm. The center wavelength interval is not less than 2nm.

3. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, The silicone rubber elastomer has a Shore hardness of A20 to A50 and an elastic modulus of 0.5MPa to 2.0MPa. The external dimensions of the flexible tactile unit are no greater than 15mm×15mm×8mm. The encapsulation depth of the seven-core optical fiber in the elastomer is 40% to 60% of the thickness of the elastomer.

4. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, The wavelength division multiplexing demodulation module has a sampling rate of not less than 1 kHz and a wavelength resolution of not more than 1 pm. In the multi-core fiber optic grating signal acquisition step, the seven grating channels are demodulated in parallel to obtain the Bragg wavelength shift of each channel, and the original wavelength shift is subjected to bandpass filtering with a bandwidth of 0.5 Hz to 50 Hz to suppress low-frequency drift and high-frequency noise.

5. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, In the temperature strain common-mode separation step, the common-mode drift component is calculated as follows: the weighted average of the wavelength drift of the seven channels is used as the temperature common-mode component, the weighting coefficient is determined by normalization based on the temperature sensitivity coefficient of each fiber core, and the temperature compensation residual does not exceed 2% of the pure strain signal amplitude.

6. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, In the step of constructing the elasticity transfer matrix, the process of constructing the transfer matrix includes: modeling the normal force as a compressive load acting uniformly on the surface of the elastic body, modeling the tangential force as a shear load acting along the plane of the surface of the elastic body, calculating the axial strain components caused by each load component at each fiber core position based on the superposition principle of elasticity, and converting the strain components into wavelength drift through the grating strain sensitivity coefficient, and assembling them to form the transfer matrix.

7. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, In the multidimensional force decoupling solution step, the weighted pseudo-inverse operation includes: constructing a diagonal weighted matrix, where each diagonal element is determined according to the signal-to-noise ratio of the corresponding channel, and the higher the signal-to-noise ratio, the greater the corresponding weight; performing singular value decomposition on the weighted transfer matrix, truncating singular value components smaller than 1% of the maximum singular value to suppress the noise amplification effect of the ill-conditioned matrix, and obtaining a regularized pseudo-inverse matrix for three-dimensional force solution.

8. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, In the tactile data fusion and output step, the contact position is estimated as follows: the six peripheral fiber cores are divided into three symmetrical fiber core pairs according to their spatial positions, the wavelength drift difference of each symmetrical fiber core pair is calculated, and the eccentricity and eccentricity direction angle of the contact force on the surface of the elastomer are determined according to the proportional relationship of the three differences, thereby estimating the contact position coordinates.

9. The flexible fiber optic grating multidimensional tactile sensing method according to claim 1, characterized in that, Before the construction step of the elasticity transfer matrix, a calibration step is also included: using a six-degree-of-freedom force sensor as a reference force source, a known force load along the three coordinate axes is applied to the flexible tactile unit, the wavelength drift response of the seven grating channels under each force load is recorded, the calibration transfer matrix is ​​obtained by fitting using the least squares method, and then fused with the elasticity theory transfer matrix to correct the model error.

10. A flexible fiber optic grating multidimensional tactile sensing system, used to implement the flexible fiber optic grating multidimensional tactile sensing method according to any one of claims 1-9, characterized in that, include: The multi-core fiber grating signal acquisition module includes a seven-core fiber encapsulated in a silicone rubber elastomer and a wavelength division multiplexing demodulation unit. The seven-core fiber includes a central fiber core and six peripheral fiber cores, and each fiber core is engraved with a Bragg grating. The wavelength division multiplexing demodulation unit is used to synchronously acquire the Bragg wavelength shift of the seven grating channels. The temperature strain common-mode separation module is used to receive the wavelength drift of the seven grating channels, extract the common-mode drift component, subtract the common-mode drift component from the original wavelength drift of each channel, and output the pure strain wavelength drift vector. The elasticity transfer matrix construction module is used to establish the transfer matrix between the pure strain wavelength drift vector and the three-dimensional force components based on the elastic body's geometric parameters and the coordinates of each fiber core position. The multidimensional force decoupling solution module is used to perform weighted pseudo-inverse operations on the transfer matrix, mapping the pure strain wavelength drift vector into a three-dimensional force vector containing normal force and biaxial tangential force. The tactile data fusion and output module is used to calculate the force vector magnitude and direction angle based on the three-dimensional force vector, estimate the contact position by combining the differential response of the peripheral fiber core grating, and encapsulate the three-dimensional force time-series curve, contact position and force vector visualization data into a structured tactile data package for output.

Images