An electronic fingerprint device based on spiral patterned tactile pixel array for augmented human-machine interactions

Abstract

Human-machine interface (HMI) devices with tactile sensing capabilities can increase both the accuracy and diversity of interaction activities. To date, the high-accuracy extraction of slippage characteristics, including both direction and displacement, remains a key challenge in HMI devices. Herein, we propose an enhanced electronic fingerprint (E-fingerprint) device that can simultaneously detect normal force, slippage direction and displacement for effective and augmented human-robot interactions. This device consists of a force-sensitive unit and a pixel array with 12×12 tactile pixels distributed in Fermat spiral pattern inspired by sunflower seeds, which highly improves the accuracy of slippage direction and displacement to 0.85° and 0.091 mm, respectively. By applying this E-fingerprint in robotic manipulation, the detected slippage can be restricted within an ideal range. A demonstration of remote writing indicates its promising performances in complicated interactive tasks. This high-accuracy E-fingerprint not only enhances the reliability of human-robot interactions but also improves interaction accuracy.

The advancement of human–machine interfaces relies on wearable tactile devices that combine flexibility with portability. Jin et al. present an electronic fingerprint device capable of simultaneously detecting normal force, slippage direction, and displacement, enabling enhanced human–robot interactions.

Introduction

With the rapid development of automated systems and artificial intelligence, human-machine interface (HMI) plays a significant role in bridging the interaction gap between human operators and intelligent robotics^1–3. Over the past decades, numerous interactive devices like touch screens, cameras, keyboards, and voice receivers have been developed to capture and process human input^4,5. However, most of these devices are bulky with complicated auxiliary equipment, lacking dexterity for conformal wearing to further extract complicated human inputs like myoelectric signals and elaborate tactile manipulating movements⁶. Due to their favorable flexibility and portability, wearable tactile devices have revolutionized human-machine interactions^7–9, which have been widely applied across the wide fields of prosthetic hand^10,11, brain-computer interface (BCI)^12,13, virtual reality (VR)/ augmented reality (AR)^14–16, robotic control^17–19 and so on. Additionally, by developing wearable tactile devices with complicated sensing capabilities, more information like multi-axis force, strain, temperature, and humidity can be simultaneously detected, improving both the accuracy and diversity of human-machine interactions^20–22.

For robust HMI applications, wearable tactile devices should reliably detect and distinguish between normal pressing forces and shear squeezing forces to accurately interpret varied user inputs. For instance, in delicate robotic manipulation tasks, stable operation of the grasped objects mainly relies on the precise monitoring of various physical parameters such as force, slippage, and so on^23,24. Researchers have proposed several solutions for HMI devices to enable the detection of both normal and shear forces^25–29. However, most works mainly focus on improving the normal force sensing performance, while both the shear force and slippage signals are indirectly achieved by either calculating multiple normal force units^30,31, using frequency-domain algorithms^32,33 or deep-learning assisted methods^34,35. So far, few studies have reported a direct strategy for precise extraction of slippage characteristics, including both direction and displacement³⁶, which are important for the quantitative analysis of slippage dynamics in robotic grasping systems. In human-machine remote interactive tasks, accurate real-time slippage signal detection can not only improve the robotic motion precision, but also enhance the adaptation to more complicated tasks^37,38. Therefore, it is highly desired to develop an advanced HMI device for robotic manipulation with high accuracy, especially in slippage detection.

High-performance slippage detection with high sensitivity and accuracy is especially important for HMI. Early detection of incipient slip provides immediate feedback that is useful for low-effort operation and fine control tasks. This is especially critical in low-force grasping, demanding high sensitivity and resolution comparable to vision-based methods^39,40. In addition, accurate slip sensing with limited interference from normal force, torque, and other environmental factors (like surface friction, morphology, and temperature) supports stable feedback control^41,42. For HMI based on shear interaction, displacement-driven operation can be more intuitive than force-based control, since users naturally associate sliding motions with cursor or robot movement in a manner similar to touchpads or joysticks^43,44.

In this study, we have developed a flexible electronic fingerprint (E-fingerprint) device with advanced tactile sensing performances, suitable for direct finger attachment in human-machine interactions. The E-fingerprint can simultaneously detect normal force, slippage direction and displacement, enabling accurate slippage detection during robotic grasping and human-machine remote interactions. The E-fingerprint consists of a force-sensitive unit and a 12 × 12 pixel array arranged in a Fermat spiral pattern inspired by sunflower seeds, serving as normal force and slippage detection modules, respectively. Spatial distribution of the pixel array is optimized to achieve high-resolution slippage detection. The normal force sensing sensitivity of our E-fingerprint is 0.36 N⁻¹ over a sensing range of 0.80 ~ 6.0 N. Notably, the E-fingerprint achieves impressive slippage detection accuracy, with directional precision of 0.85° and displacement resolution of 0.091 mm. The E-fingerprint mounted onto a robotic hand successfully detects and prevents slippage during object-grasping tasks. We further demonstrate remote robotic control of a letter-writing task using the E-fingerprint device worn on a finger, showcasing its great potential for HMI tasks. Therefore, this improved E-fingerprint device has great potential for future human-machine interactions.

Results

Concept and structural design of E-finger print device

In order to overcome the challenge in precise measurement of slippage direction and displacement, an overlapped fingerprint-shaped device with a spiral-patterned tactile pixel array is proposed. Figure 1a illustrates its operating principle when applied in human-robotic interactions. In detail, when mounted on a hand finger (left part in Fig. 1a), this device acts as a finger-imitating control interface. The operator can simply press or slide the top layer of the device along arbitrary directions to generate interactive information of normal forces and slippages. Subsequently, an anti-crosstalk circuit collects these tactile signals, which are further analyzed on the upper computer to extract tactile features. Once the force or slippage values accrue to a critical threshold, the robotic hand receives the motion instructions and completes remote writing tasks, which are precisely monitored and controlled by human beings. Moreover, the device can also be integrated onto a robotic hand to improve grasping stability due to its ability to accurately monitor the slippage signals. Such a tactile sensing E-fingerprint can effectively improve the reliability of interaction between humans and robots, showing promising applications in robotic delicate manipulation, remote medicine, virtual reality entertainment, and so on.

Fig. 1. Schematic illustration shows the application and structural design of E-fingerprint device.

a The working principle of E-fingerprint. The transmitted tactile signals can be utilized for application scenarios like robotic remote control and robotic grasping. b Structural design of E-fingerprint. The force sensing unit is designed using a hollow structure with a kirigami pattern, and the 12 × 12 tactile pixel array is arranged in a Fermat spiral pattern that inspired by the distribution of sunflower seeds. Image of the sunflower adapts from Pixabay user “BRBurton23” (licensed under CC0, https://pixabay.com/photos/sunflower-yellow-spiral-plant-1159732/). c Photograph of the E-fingerprint prototype. Scale bar: 5 mm.

Figure 1b shows the overall structure design of E-fingerprint, it mainly consists of three layers from top to bottom: an encapsulation layer, a force sensing layer, and a slip sensing layer. The effective sensing area of the device is a circle with a radius of 7 mm, which is suitable for mounting on a human finger. The encapsulation layer is made of polydimethylsiloxane (PDMS) with a fingerprint imitating appearance, as shown in Supplementary Fig. 3. The laser-engraved grooves in the central region is designed based on human fingerprint, which can increase the friction force⁴⁵. And the curved hollow structure on the periphery uses a vortex-like design that also mimics a human fingerprint, which endows the central part with excellent flexibility to move in arbitrary directions. And the hollow structure outside the sensing area can improve the flexibility of this device to be mounted on a curved surface. Using a similar structure design to the top, the force-sensing layer replaces the central part with a force-sensitive resistive rubber sandwiched by two electrodes made of conductive nanocomposites. Specifically, the force-sensitive rubber has a kirigami hollow design⁴⁶ using micro-pillars with a height of 150 μm decorated on the surface⁴⁷, which can greatly improve the structure’s stretchability and enlarge the force sensing sensitivity. All the contact regions between the encapsulation and force-sensing layers are firmly bonded, while the slip-sensing layer only connects with the force-sensing layer in the peripheral area. Inspired by the layered structure of subcutaneous tissue in human skin that can slide against each other⁴⁸, the force-sensitive unit in the middle layer can also slide on the bottom layer due to the curved hollow structure on periphery (Supplementary Movie 1). A pixel array of 144 units with a small spatial resolution of 1 mm acts as a digital interface measuring both the slippage direction and displacement. Each pixel is composed of a pair of electrode pads and a resistive composite with a radius of 500 μm. Sunflower seeds are arranged based on Fermat spiral lines, which avoids the imbalance problem of overcrowding at the center and sparseness at the edges, achieving the maximum utilization within a limited circular area and Maximizing the efficiency of photosynthesis^30,49. Inspired by this distribution, a group of 12 pixels are aligned in a shape of Fermat spiral (Supplementary Fig. 4) and connected with electrodes, which provides higher measuring resolution compared with traditional square or circular shaped array⁵⁰. The detailed manufacturing process of the device is indicated in Supplementary Figs. 5 and 6, the fabricated device prototype is shown in Fig. 1c. The shear strength of bonding interfaces with heterogeneous materials in the device are shown in Supplementary Fig. 7, which proves good bonding between the interfaces. From Supplementary Movie 1, we can see that the E-fingerprint can bear large deformation, indicating its great flexibility to be mounted on curved surfaces.

Working principle and optimization of pixel array distribution

Based on piezoresistive effect, a resistive signal can be generated by pressing the force-sensitive unit, as shown in Fig. 2a. The micro-pillars with a diameter of 100 μm and a height of 150 μm are fabricated by laser engraving, and the morphologies are shown in Supplementary Fig. 8. Besides, the kirigami-patterned hollow structure with a width of 300 μm can reduce the resistance change caused by shear strain. The normal force sensing process can be divided into two stages (Supplementary Fig. 9). When the top surface of the device is subjected to normal force, the contact area between the micro-pillars and conductive electrodes first changes upon the external force, which improves the force sensing sensitivity when considering the contact resistance (R_c)⁵¹. Furthermore, the bulk resistance (R_i) of conductive rubber decreases since the effective conducting path between the conductive particles shrinks⁵². After activating the force-sensitive unit, the external load further transmits to the pixel array in the slip sensing layer downward, the working mechanism of which is shown in Fig. 2b. For each sensing pixel, a circular unit made of conductive nanocomposites is suspended upon a pair of electrodes with a vertical gap of 100 μm, forming an initially disconnected circuit. Once the normal force is applied, the conductive unit will contact the electrodes and therefore connect the circuit. Thus, each pixel acts as an electrical switch (S_p) that uses the on-off states to generate the 1-0 logical signal (R_p). As the 12×12 pixels form a sensing array, more tactile features like slippage can be extracted by calculating the force center, as shown in Fig. 2c. When applied by normal force (left part), the pixels below the force sensing unit will be activated, reflecting the projection area of the external load. And if a slippage is further loaded, the top layer can slide on the bottom pixel array due to the curved hollow structure on periphery, and the force projection area also moves along the slippage direction correspondingly (middle part). The activated pixels are highlighted in real-time on a software interface to visualize the area projection of external load, as shown in Supplementary Movie 1. In addition, an estimation method to calculate the coordinate value of the force center is developed, as explained in Supplementary Note 1. By calculating the average coordinate of all activated pixels, both the angle (θ) and offset (d) of the force center can be estimated, which represent the slippage direction and displacement, respectively (right enlarged part). Therefore, the developed device can simultaneously detect multiple tactile signals, including normal force and slippage. Detailed slippage information can also be further extracted precisely.

Fig. 2. Working principle of E-fingerprint device for tactile sensing.

Schematic illustration showing the electrical circuit and working principle of a force-sensitive unit and b slippage-sensitive pixel. c Working principle of pixel array for slippage detection. The projection area of the force center when only applied by normal force (left part), and when applied by both normal force and slippage (middle part). In the right enlarged part, p and p’ are the real and estimated central point of the projection area, o is the original point, d and θ are the slippage offset and angle. d The distribution pattern of pixel array. φ₁ and φ₂ are characteristic parameters of Fermat spiral lines. The normalized spatial dispersion of Fermat spiral patterned pixel arrays with the characteristic parameters of φ₁/φ₂ lower (e) and higher (f) than 0.618. g The spatial dispersion and average error of spiral patterns with various spatial angle ratios. Error bars represent standard deviation (s.d.) with the experiment repeated for 3 times (n = 3).

The detection accuracy of slippage mainly depends on the distribution of tactile pixels. In a traditional tactile sensing array, a rectangular arrangement of numerous pixels is generally adopted for the ease of row and column signal scanning⁵³. When considering the consistency for arbitrary directions, only the pixels in the inscribed circle of the rectangular area are effective for slippage detecting, which means at least 21% of the pixels at the corners will be wasted. A circular-shaped array of pixels is suitable for slippage detection, but the distribution of the pixels can still be optimized to improve the sensing performance. Herein, a sensing array with 144 pixels distributed with a Fermat spiral pattern is proposed, as shown in Fig. 2d. A group of 12 pixels are connected in series by a row electrode (red line), while the column electrode (blue line) aligns adjacent pixels selected from 12 pixel groups in the opposite direction. Both the electrodes are designed with a Fermat spiral pattern, and the detailed equation of these spiral lines are explained in Supplementary Note 1, Supplementary Fig. 1 and Supplementary Fig. 2. By adjusting the characteristic parameter, the rotation angle of spiral line will also change accordingly. To further demonstrate the superiority of this array, a quantitative evaluation method analyzing the average uncertainty, namely the spatial dispersion level of tactile pixel distribution, is discussed in Supplementary Note 2. A higher spatial dispersion level represents a more uniformly distributed pixel array, which contributes to a higher resolution for slippage detection. For distribution patterns like square, rhombus, circle and Fermat spiral (Supplementary Fig. 10(a)), the calculated spatial dispersion level is illustrated in Supplementary Fig. 10(b). For all the distribution patterns, the spatial dispersion gradually increases when the space is divided into more segments. For a Fermat spiral line, the incremental value of the angle between adjacent pixels, namely the spatial angle in row and column electrode is defined as φ₁ and φ₂, respectively. By continuously changing the design parameters, the spatial dispersion achieves the highest value when the spatial angle ratio of the spiral lines φ₁/φ₂ is approximately 0.618 (Fig. 2e–g), which is superior to traditional patterns (Supplementary Fig. 10b) except the circle type. More quantitative comparation between Fermat-spiral and other patterned arrays are listed in the Supplementary Table 1, which shows the superiority of our design in detail. And some traditional circular pixel arrays are illustrated in Supplementary Fig. 11. The square patterned pixels arranged in a circular area can be easily mapped by row-column scanning, but the shape cutting design may provide redundant information and unevenness on boundaries. Besides, the spider-web-like pixel array is suitable for circle design, but the resolution of slippage direction on periphery is lower than that in center. The close-arranged concentric circles pixel array has comparative spatial dispersion with our Fermat spiral design, but it may utilize complicated structure or signal processing algorithm to map the pixel array^54–56, which may also increase the size of the device. Thus, the spiral line with a φ₁/φ₂ of 0.618 is chosen to design the distribution pattern with the highest spatial dispersion level, which represents a higher slippage detection resolution.

Sensing performances of E-fingerprint

To evaluate the electrical performances of the developed E-fingerprint device, a series of experiments are conducted to characterize their sensing performance. For normal force-sensitive unit, the force sensitivity is characterized as 0.36 N^-1 in a force range of 0.80 ~ 6.0 N, which satisfies the demand for press controlling and daily grasping tasks⁵⁷, as shown in Fig. 3a. The sensing range is mainly determined by the intrinsic viscoelasticity of conductive rubber. The shift in the dominant conductive component introduces increased variability in the measured resistance, leading to the observed scatter.

Fig. 3. Electrical characterization of E-fingerprint for tactile sensing.

a Normal force sensitivity (S) of the force-sensitive unit. Error bars represent standard deviation (s.d.) with the experiment repeated for 3 times (n = 3). b Dynamic response of the force-sensitive unit. T_L and T_UL are the loading and unloading times. c Activation (F_on) and inactivation (F_off) force threshold of each pixel unit. d Dynamic response of each pixel unit. e Comparison between the true and experimental slippage of varied angles (0–360°) and amplitudes (2-6mm). Error bars represent standard deviation (s.d.) with the experiment repeated 5 times (n = 5). f Calibration results of the slippage direction from 0° to 330° with an interval of 30°. g Deviation and resolution of slippage direction. h Calibration results of the slippage distance with an interval of 2 mm. i Deviation of the slippage distance. j Comparison of shear angle resolution in slippage detection between this work and other reported works^{23,26–29,37,38}. k Sensing performance of the device when simultaneously applied by both normal force and slippage.

When a normal load changes between 10% and 90% of the full sensing range, the loading time (T_L) and unloading time (T_UL) are tested to be 360 ms and 120 ms, respectively (Fig. 3b). Besides, the stability and repeatability of the force sensing unit are characterized using forces with varying amplitude and frequency (Supplementary Fig. 12a and b), and approximately 10000 repeatable loading cycles (Supplementary Fig. 12c and d). Finally, the hysteresis test is also conducted (Supplementary Fig. 12e). The relatively large hysteresis is mainly caused by the viscoelastic nature of elastomer and the stress relaxation phenomenon, which can be eliminated by signal self-compensation method. To evaluate the friction stability on the normal force sensing unit covered by fingerprint-engraved PDMS layer, a friction endurance test⁵⁸ is conducted (Supplementary Fig. 13), and the calibration curves of the sensing unit is similar despite 1000 cycles of friction load.

For tactile pixel sensing array, the performance of each pixel is of great importance. As shown in Fig. 3c, when a time-varying force is applied to the pixel, it rapidly creates a contact change between the resistive pixel unit and electrode pads. To prevent signal interference, a voltage larger than 90% of the maximum value is defined as the stable contact state, which specifies the turn-on and turn-off force threshold values to be 0.71 and 0.52 N, respectively. The slight difference between the loading and unloading processes is due to stress relaxation in the resistive nanocomposite, which reduces the force threshold during unloading⁵⁹. In Fig. 3d, when repeated instantaneous forces are applied to the pixel unit, the measured signal response rapidly in less than 100 ms, with consistent peak values between each loading group. Moreover, for different pixel units, the response curves under the same on-off loading tests also remain consistent (Supplementary Fig. 14). And a repeatability test of 10000 cycles of slippage is also conducted (Supplementary Fig. 15), which proves good durability for the pixel sensing array.

To further evaluate the slippage detection performance of the E-fingerprint, the force center estimation experiments are conducted. First, a normal force of 1.0 N is applied to activate the pixel unit. The top layer is then subjected to an initial slippage of 2.0 mm along the angle of 0° in the horizontal polar coordinate, subsequently. The force center can be estimated based on the method discussed above. After that, the slippage angle gradually changes from 0° to 330° at an increment of 30°, and the slippage displacement also increases to 4.0 and 6.0 mm. As compared in Fig. 3e, the force centers estimated by the E-fingerprint are in good agreement with the true value in all directions, indicating the improved capability of this device in detecting omnidirectional slippage. A more detailed calibration of the slippage angle from 40° to 50° with an increment of 1° is also provided in Supplementary Fig. 16 to verify the accuracy in angle estimation. Figure 3f and Fig. 3h shows the calibration results of the estimated direction and distance compared to the true value. The deviation values of the estimated slippage are shown in Fig. 3g and Fig. 3i, in which the average deviation values of slippage direction and displacement are characterized to be 0.85° and 0.091 mm, respectively. Such a high slippage detecting accuracy is due to the unique pixel distribution in the Fermat spiral pattern, which can be further improved by increasing the number of pixels. To the best of our knowledge, the slippage detection accuracy of this E-fingerprint surpasses that of the most advanced tactile sensing devices recently reported (Supplementary Table 2 and Fig. 3j).

In Fig. 3k, the normal force and slippage are sequentially applied to verify the low coupling error. The slippage signals with different displacements of 4.0 mm and 4.5 mm can be clearly distinguished based on the measured signals. The slippage signal may fluctuate during the loading or unloading process of the normal force, but it returns to a stable state when the force signal becomes steady. The changing of the force signal during the slippage loading process is normal since the slippage is generated by both normal pressure and lateral shift. A more detailed coupling test between the normal force and slippage sensing is provided in Supplementary Fig. 17, which proves low machinal interference between different sensing layers. Actually, the shear force can also be evaluated based on the force analysis of the central part, in which the peripheral curved hollow structures can be simplified as Springs (Supplementary Fig. 18). The calibrating results of shear force under different slippage and normal forces is provided in Supplementary Fig. 18b and c. The slippage deforms the peripheral structures to change the stretched or compressed forces, and the normal force influences the friction coefficient of the contact surface between the central part and the sliding substrate.

In short, the developed E-fingerprint can simultaneously detect normal force and slippage with little interference, and it also has enhanced capabilities to extract slippage characteristics such as direction and displacement with high accuracy, indicating its great potential to be utilized as a human-machine interface.

Applications of E-fingerprint

Since the E-fingerprint has enhanced slippage detection capabilities, it’s beneficial for applications such as dexterous robotic manipulation. This flexible E-fingerprint can be utilized as a tactile sensor and mounted on a robotic hand to enhance the grasping stability, and the overall signal transmitting procedure is illustrated in Fig. 4a and b. Specifically, the robotic hand mounted with our E-fingerprint is controlled to grasp a test tube. By continuously injecting water into the test tube, both normal force and slippage signals are monitored through the anti-crosstalk scanning circuit. The signals are then transmitted to the host computer, which can analyze tactile features and send movement instructions to the end-effector to prevent slippage. The real-time grasping photos and the tactile messages extracted during the whole process are illustrated in Fig. 4c. The test tube is grasped with a grasping force of 0.15 N (left part), and the initial slippage caused by gravity is about 0.5 mm with an angle of 120°. When the water is continuously injected, the estimated slippage displacement remains stable until gravity exceeds the friction force at 3.0 s. Once the detected slippage displacement exceeds the threshold value of 2.0 mm at 3.3 s (middle part), the robotic hand rapidly increases the grasping force to 1.1 N at 3.7 s to prevent further falling, which restricts the total slippage displacement within 3.0 mm (right part). The grasping force is monitored by the sensing device in real-time, which is also sent to the robotic arm to prevent object from being damaged. The estimated slippage angle fluctuates before the increase of the normal force due to the impact between the water and the test tube. Since the slippage measurement has low coupling effect with the normal force sensing (Supplementary Fig. 17), and the slippage prevention is directly based on the measured slippage signal, the results remain effective and stable despite the monitored hysteresis of the normal force (Supplementary Fig. 12e). Compared to other reported works that using normal sensing matrix to detect slippage or shear information^26,60–63, as indicated in Supplementary Table 3, our device is compact, flexible and easy-fabricated. And it can achieve high resolution of slippage detection using only spatial information without a data-intensive signal training algorithm or complex structural model establishment, which is more convenient to expand and integrate on robotic actuators.

Fig. 4. Slippage prevention during robotic grasping process.

a Schematic diagram of a robot integrated with the E-fingerprint. b Photograph of the E-fingerprint mounted on the gripper. Scale bar: 25 mm. c The force and slippage signals during robotic grasping when water is continuously added into the test tube. Stable grasping before slippage (left part). The slippage happens with a displacement exceeds the threshold value of 2 mm (middle part). Stable grasping again after controlling the slippage displacement within a total value of 3 mm (right part).

During the real-time monitoring process, the shape of the activated pixels is not exactly the projection area of the force-sensitive unit due to manufacturing errors (Fig. 4c). An iterative algorithm will be introduced to improve the weights of the pixels in adjacent to the estimated slippage center, which can improve the accuracy of slippage estimation. Besides, we have noticed that the total slippage consists of two components: one is the internal slippage, which is between the force sensing layer and the slippage sensing layer; the other is the external slippage, which is between the grasped object and the encapsulation layer. And the relationship between these two slippages can be demonstrated in Supplementary Fig. 18. Thanks to the design of top laser-engraved grooves that increase the friction force, the internal slippage will happen earlier than the external slippage. Although the external slippage is not directly measured, the internal slippage can also represent the contact state between the mounting substrate and the object. Since the internal slippage is more likely to happen, the tiny slippage can be controlled earlier before it increases to a larger value. A more detailed demonstration of slippage prevention process is shown in Supplementary Movie 2.

The E-fingerprint also shows potential applications in robotic remote control. As illustrated in Fig. 5a, the device is mounted on the index finger pulp, which can be easily manipulated by the thumb to generate normal force and slippage signals. The tactile signals are transmitted to the robotic arm to subsequently perform the corresponding interactive movement. A mapping relationship between the finger actions and robot motions is established (Fig. 5b), in which a sustained press controls the opening state of grasping jaws, and a rapid press can switch the current moving planes. Besides, slippage in a certain direction drives the gripper to move accordingly, and a larger slippage displacement corresponds to a faster moving speed. This demonstration underscores how the tactile device is utilized as a human-machine interface with high control precision for delicate daily tasks. Figure 5c shows a demonstration of the E-fingerprint applicated in robotic arm remote moving and writing. When the manipulator applies zigzag slippage signals to the tactile device from 20 s to 45 s, the black pen successfully writes a letter of “Z”, indicating a good controllability of the slippage direction. Then, the moving plane is switched from the horizontal plane to the vertical one by fast pressing in less than 1.0 s. The robotic hand is lifted and moves towards the pen cap. In this process (from 45 s to 93 s), a relatively large displacement of the slippage is applied, ensuring the end effector moves at a faster speed of 6.0 mm/s before approaching the target position. Once the pen hovers above the cap, the robotic hand is delicately adjusted by repeatedly applying slippage and rapid press signals, which help insert the pen into the cap (from 93 s to 150 s). Finally, a sustained press signal lasting more than 2.0 s is applied to loosen the grasping jaws (150 s to 180 s), indicating that the grasping mission is completed. Although there exist some variabilities in the normal force signal, both the sustained and rapid force can still be clearly distinguished, and the repeatable and fast change of the normal force during the positioning process is also distinct from each other, which proves its effectiveness in HMI. A more detailed demonstration of the remote manipulation process can be achieved in Supplementary Movie 3.

Fig. 5. Robot remote control application using E-fingerprint device.

a Schematic diagram of human-machine interaction. b Mapping relationships between finger action and robot motion. Sustained pressing controls the robotic gripper's grasping state. The rapid press can switch the motion plane. The slippage signal forces the gripper to move in a certain direction. c The force and slippage signals during robotic remote delicate manipulation. Writing the letter of “Z” (20 s to 45 s). Switching the moving plane and moving the gripper quickly (45 s to 93 s). Delicately adjusting the position of the gripper (93 s to 150 s). Releasing the gripper (150 s to 180 s).

Discussion

In conclusion, we have developed an enhanced flexible E-fingerprint device that can simultaneously detect normal force, slippage direction, and displacement, which can be utilized as a human-machine interface with high controlling accuracy, especially in the shear direction. It optimizes the arrangement of 12 × 12 tactile pixel array in a Fermat spiral pattern with a spatial resolution of 1 mm, which highly improves the detection resolution of both slippage direction and displacement to a competitive level of 0.85° and 0.091 mm, respectively. A hardware anti-crosstalk circuit and a real-time display interface are also established to improve the instantaneity of the detected tactile signals. By applying this device in the robotic grasping task, the unexpected slippage value can be precisely quantified and restricted within an ideal range of 3.0 mm. A complex manipulation of robotic remote control and writing is also conducted, indicating its promising performance as a human-machine interface, especially for slippage-based interactive movements with high accuracy. Besides, a comprehensive comparison between our work and conventional 3-dimensional distributed sensors is shown in Supplementary Table 4, which further proves the improvement of this work. Therefore, the proposed tactile device can improve interactive precision and enrich the human-robot interface, demonstrating enormous potential for the future development of superior HMIs.

Methods

Material and fabrication of E-fingerprint device

The materials of encapsulation layer and support layer are Polydimethylsiloxane (PDMS, Sylgard 184) with a mass ratio of 10:1. The force unit uses commercially available conductive rubber (INASTOMER) as the sensitive material, which is sandwiched by electrodes made by a composite of silicone rubber (SR, GD401, Zhonghao Chenguang Research Institute of Chemical Industry) and silver nanoflakes (AgNFs, Sigma Aldrich) with a mass ratio of 1:2. The sensing material for the pixel unit is prepared by mixing silicone rubber, Graphene (diameter <10 μm, The Sixth Element Materials Technology Co., Ltd), carbon nanotube (Nanjing XFNANO Materials Tech Co., Ltd) and a solvent of Tetrahydrofuran (THF) with a mass ratio of 1:4%:2%:1, which has been reported in out former works⁶⁴. Different layers of the E-fingerprint device are fabricated separately. The encapsulation layer is mainly cut and engraved by a laser cutting machine (V2000, Guangzhou HZZ E-photo Technology Co., Ltd), which is also utilized to prepare the support layers in the force sensing layer and slip sensing layer. The AgNFs/SR composite electrodes are screen-printed and heated at 80 °C for 2 h, which are further connected with the sensing unit and the support layer using a conductive silver paste (FSR-8000B, Shenzhen Sunflower Electronic Materials Co., Ltd) to form the force sensing layer. The substrate material of the bottom layer is polyimide (PI), with Cu electrodes fabricated by sputter coating and etching. The pixel array layer is also fabricated by screen-printing the sensitive material on PDMS and heating at 80 °C for 2 h. In this case, silicone rubber (GD401) is used as the adhesion agent to bond the support layer, the pixel array, and the PI substrate to form the slip-sensing layer. Finally, all these layers are combined with the silicone rubber adhesion agent to complete fabrication of the E-fingerprint device.

Electrical hardware of the circuit

The schematic diagram of the scanning circuit for anti-crosstalk signal acquisition is shown in Supplementary Fig. 19. The system and peripheral circuits are built on a microcontroller chip (STM32F407), providing logic control for resistance network sampling. An 8.0 MHz passive crystal oscillator (YSX530GA) is selected as the high-speed clock source of the system. As for the system power supply module, the LDO chip (SPX 1117) is selected to ensure low ripple on the 3.3 V DC supply. To broaden the range of resistance acquisition, a digital potentiometer (MCP41100) is used to adjust the reference voltage. The voltage follower circuit is designed as a buffer for electrical isolation. The CMOS operational amplifier (OPA2350) is selected to realize resistance measurement and signal amplification, as well as to minimize the crosstalk noise. The sampling frequency is about 80 Hz for acquiring 144-pixel array. The fabricated printed circuit board (PCB) is also shown in Supplementary Fig. 19b. All components are soldered on the FPCB using a low-temperature soldering paste.

Signal anti-crosstalk mechanism and results analysis

The detailed signal anti-crosstalk method is shown in Supplementary Fig. 20. The outside crosstalk can be eliminated by the equipotential method, and the inside crosstalk can be minimized by optimizing the circuit. The simulation results of normal circuit without anti-crosstalk design and simplified model of our circuit is shown in Supplementary Fig. 21. The simulation model is established in the Software of LT Spice. The real-time signal shows the crosstalk influences from adjacent units is shown in Supplementary Fig. 22.

Characterization and measurements

The performance characterization experiments are conducted on a multi-axis force platform, as shown in Supplementary Fig. 23. Normal forces applied to the tactile device are characterized by a high-precision load cell with a sensing range of 20 N and a resolution of 0.001 N. The load cell is mounted on a test bench (TA Instruments, USA) that can move in the Z-axis with a resolution of 0.01 mm. For slippage detection tests, the tactile device is first applied with a normal force, and it is mounted on a manual rotation platform (PT R60, PDV) to change the slip direction, which can be adjusted from 0 ° to 360 ° with a resolution of 55 °. And the rotation stage is mounted on a horizontal moving stage (PT110-13-34, PDV) for X-axis movement, which can provide a maximum slip displacement of 13 mm with a resolution of 0.01 mm. All the force and slippage signals are collected by the designed anti-crosstalk circuit and transmitted to the upper computer.

Robotic setup for applications

The robotic system includes a robotic arm (TM5-900, Techman Robot Co., Ltd.) with a robotic hand (2F-85, Robotiq Co., Ltd.) with two gripping jaws. The robotic arm has a maximum load of 4.0 kg and a repeat positioning accuracy of 0.03 mm. And the repeat positioning accuracy of the robotic hand is 0.05 mm, with a maximum clamping force of 235 N.

Author contributions

J.J., C.M. and Y.W. conceived the idea and developed the methodology. C.M. established the theoretical models. C.M., W.L. and Y.W. performed the investigation. C.M. conducted the characterization experiments. J.J. and C.M. carried out the experimental applications. Y.W. and D.M. supervised the research. J.J. wrote the original draft of the manuscript. Y.W., Y.L. and D.M. reviewed and edited the manuscript. All authors discussed and approved the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work is supported by the following funding: National Natural Science Foundation of China (No. 52475572, 52175522), Zhejiang Provincial Natural Science Foundation of China (No. LZ24E050003), Key Research and Development Program of Zhejiang Province (2025C01003), Guizhou Provincial Science and Technology Projects (No. [2025]033).

Data availability

All relevant data generated in this study are provided in the Source Data file. More detailed protocols, calculations and analyses are available from the corresponding authors upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-73973-6.