Fully 3D Printed Flexible, Conformal and Multi-directional Tactile Sensor with Integrated Biomimetic and Auxetic Structure

doi:10.21203/rs.3.rs-877986/v2

Download PDF

Article

Fully 3D Printed Flexible, Conformal and Multi-directional Tactile Sensor with Integrated Biomimetic and Auxetic Structure

https://doi.org/10.21203/rs.3.rs-877986/v2

This work is licensed under a CC BY 4.0 License

Journal Publication

published 14 Nov, 2023

Read the published version in Communications Engineering →

Version 2

posted

You are reading this latest preprint version

Tactile sensors are instrumental for developing the next generation of biologically inspired robotic prostheses with tactile feedback capability. However, current sensing technology is still less than ideal either in terms of sensitivity under high pressure or compliance with uneven working surfaces. Also, the fabrication of tactile sensors often requires the use of highly sophisticated and costly manufacturing processes further limiting the widespread application of the technology. Here, we challenge the current perspective and propose the use of an in-house 3D printing system to develop a new conformal tactile sensor with enhanced sensing performance. The ability of the sensor to detect multi-directional stimuli is achieved through the integration of the auxetic structure and interlocking features. The unique design of our sensor allows for an extended sensing range (from 0.1 to 0.26 MPa) whilst providing sensitivity on both normal and shear directions at 0.63 KPa^− 1 and 0.92 N^− 1, respectively. This is further complemented by capacity of the sensor to detect small temperature variations between 40 and 90°C. To demonstrate the feasibility of our approach, the tactile sensor is printed in situ on the fingertip of an anthropomorphic robotic hand, the proximal femur head and lumbar vertebra. The results suggest that it is possible to gain sensorimotor control and temperature sensing ability in artificial upper limbs whilst monitoring the bone-on-bone load, thus opening the door to a new generation of tactile sensors with novel auxetic structure design and enhanced performance for application in human prosthetics.

Over the past decade, great attention has been dedicated to the development of wearable, flexible tactile sensors and electronic skin for application in prosthetics, robotics and other healthcare devices ^4,5,7−15. These applications often require the use of tactile sensors able to detect multi-directional stimuli. To that end, different strategies have been proposed including the use of advanced composite materials, multi-layered biomimetic or hierarchical structures ^1,5,11,15,16. However, the fabrication of high-performance sensors with conventional manufacturing techniques such as molding, photolithography and etching ^17–20 remain challenging, mainly due to the complex and time consuming fabrication processes and material preparation stages. In terms of design, sensors often display simple, planar geometries which are incompatible with the majority of working surfaces (irregular or uneven), thus limiting their performance and range of applications ^21–23. Soft optical sensors with tactile feedback ability, such as the Gelsight ²⁴ and Tactip series sensor ²⁵, have also been proposed by other research groups and integrated into robots to achieve sensorimotor control. However, the use of such sensors remains hindered by the large space required to accommodate the camera and other supporting components.

In an attempt to improve the sensitivity and linear sensing range of tactile sensors, several groups have reported the combination of highly conductive composite materials such as metallic particles, nanowires ¹¹, carbon nanotubes (CNT) ²⁶ and graphene palate ²⁷ with an insulation matrix. Carbon nanotubes CNT/polymer composites displaying high axial conductivity are particularly attractive for the generation of sensors with dielectric layers. When compared to similar quantities of carbon black or graphene, the addition of CNT to polymer matrices induce a substantiality higher piezoresistive response with minimal effect on the mechanical properties of the composite ^28–30. Other approaches based on volatile or water soluble micro-particles have also been employed to generate sponges or porous material ^13,27 with higher self-contact area and larger linear sensing range. However, the process for producing these porous materials is time-consuming and technically complicated. Also, the stability and homogeneity of the material cannot be precisely controlled thus negatively affecting the final structure and performance of the sensor. Alternative routes toward the design of sensors with improved sensitivity and linear sensing range include the use of multiscale hierarchical structures ^2,10,11, biomimetic interlocked structures ^5,31−33, and mechanosensory hair or even crack-like structures ^34,35. The integration of microspheres with nanomaterials has also been used as a potential avenue to obtain hierarchical structures with enhanced piezoresistive properties ¹. In recent studies ^2,5,15, a series of interlocked micro-dome arrays emulating intermediate ridges in the human skin have been proposed as ideal systems to amplify signal responses and differentiate between multiple stimuli directions. Because of their biomimetic structure these systems also found application as optical-based tactile sensors to improve the capability of discriminating fine features ²⁵. In a similar vein but using multilayer or hierarchical hair-like interlocked geometries ^11,14, other researchers have demonstrated the ability to produce structures with improved sensitivity and extra low detection threshold of 0.6 Pa. However, these structural features require the use of sophisticated and time-consuming fabrication processes. The pressure sensor that can be 3D printed or fast-prototyped was developed by Guo et al.³⁶ while a simple coil structure was adopted and limit the sensing performance. The expensive Ag-silicone rubber material was used for printing which limit its practical applications in industry. The sensing area of this 3D printed was also small and less than 10 mm². In spite significant progress, the reality is that most of the sensors reported above have not yet found a practical application in the industry. To accelerate the translation of the technology from the bench to the bed side and deliver the next generation of prosthetics, new strategies must be developed for the fabrication of pressure sensors with: (1) less complex and more affordable manufacturing processes; (2) similar or larger sensitivity and linear sensing range compared with published sensors; (3) a stronger, more stable and durable interface with working surface; (4) a customized sensing area ranging from millimetre to centimetre in size.

In this paper, we present a novel flexible tactile sensor that can detect both contact pressure and environmental temperature. For the first time, a biomimetic and auxetic structure is integrated in the sensor providing the ability to detect multidirectional stimuli. Using 3D printing technology we show the feasibility of our approach to quickly generate low-cost conformal sensors directly on top of uneven working surfaces. We demonstrate that the in situ printed sensors can be effectively used to gain sensorimotor control and temperature sensing ability in a biomimetic hand, and to provide the accurate bone-on-bone load monitoring in human joints (i.e. vertebra and femur). When compared to sensors with planar structure in this study and others in the literature ^13,37−40, our unique sensor design with integrated biomimetic interlocked and auxetic structure provides significant benefits in terms of sensing performance, namely: 1) larger linear sensing range due to the negative Poisson’s ratio; 3) higher sensitivity at low pressures (10 KPa); and 3) easy detection of the stimuli direction. Besides the outstanding pressure sensing performance, the sensor can also respond to temperature variations between 40 to 90°C with a sensitivity of 0.27% °C^− 1. More than simply overcoming the limitations of the current technology, this study aims to pave the way towards the design and fabrication of a new generation of tactile sensors with user-defined auxetic features suitable for applications in robotic/prosthetic hands and in pressure/temperature monitoring in impaired human joints.

The tactile sensor – Design and 3D printing

The layout of the flexible tactile sensor array consisting of 121 sensing elements is depicted in Fig. 1a-e. The new structure contains the upper and lower papilla-auxetic dielectric layers and is sandwiched between two flexible electrode layers. The sensing area and distribution density of the tactile elements are comparable with slowly adapting type I (SAI) and fast adapting type I (FAI) mechanoreceptor to ensure a close match between the sensing capability of the proposed sensor and that of a human subject ⁴¹. The biomimetic interlock structure is integrated to enhance the sensitivity and provide capability of discriminating between different directions of the external stimuli. Our results show that the piezoresistive sensing elements (located at the position of each small papilla) placed on the side of the stimuli experience substantially larger reduction of resistance compared to those placed on the opposite side. This can be associated with the geometric features of the lower large papilla and the anisotropic deformation of the upper dielectric layer of the small papilla. The direction of the stimuli can be intuitively differentiated based on each large papilla surrounded by four small papillae above its four corners as highlighted with in Figure. 1f (red squares). The two small papillae on the side of the stimuli experience larger current increments than the other two on the opposite side. Therefore, the pressure mapping on the tactile sensor array, originated by the anisotropic distribution of resistance around the lower papilla, provided the ability to differentiate between the different directions of the stimuli. On the other hand, the single sensing element was not able to provide the bulk information on the directions of the external stimuli.

A Graphene/CNT/Silicon rubber composite is used as the printing material for fabricating the proposed tactile sensor. The graphene palate and multi-wall CNT are evenly distributed within the silicon matrix (Fig. 2a), providing piezoresistive and thermoresistive properties to the sensor. The CNT to graphene weight ratio was optimized to maximize the temperature resistive coefficient (TRC). The conductive micro-copper-silicone composite is employed to print the electrode. Printability of the developed composite materials is determined through oscillatory rheological measurements, in particular by measuring the variation of viscosity and shear stress as a function of shear rate (see Supplementary Figs. 1 and 2). Both composite materials exhibit a shear thinning behavior typical of non-Newtonian systems with extrusion being enabled by a reduction in viscosity as a consequence of increased shear stresses.

The sensor is manufactured using a 3D printing platform developed in-house and based on the concept of Fused Deposition Modelling (FDM). The printing process of the sensor from the modeling stage to the conformal 3D printing onto the freeform surfaces is illustrated in Fig. 2b. Once, the target surface is identified and modelled, the object is discretized with a fine parametric mesh using the CAE software Abaqus (Dassault Systèmes Simulia Corp, Providence, RI). Then, the coordinates of the node close to the printing paths conformal with the desired surfaces are extracted. Finally, a G code containing the numerical coordinates of the nodes (i.e. printing pathway) is generated and sent to the printer allowing the sensor to be directly fabricated onto the working surfaces (see Figs. 2(c) to 2(f) ). The upper and lower layers of the sensor are printed separately. The former is printed onto the ‘negative surface’ and peeled off to attach with the lower layer on the working surface. The negative surface is modelled conformally to the positive working surface and then printed through Stereolithography (SLA). The G codes for printing the sensors onto different positive working surfaces and negative surfaces are presented in the supplementary material.

Figure. 2(g) shows the response of the sensor under the contact with the fingertip, in the trials the fingertip pressed the tactile sensor, the direction of the shear stimuli was indicated by the arrows in yellow. The distribution of the contact pressure initiated by the external stimuli are shown through customized graphic user-interface (GUI). It can be seen that the sensing elements on the sides of the external stimuli experienced larger resistance variation.

Auxetic structure optimization of the tactile sensor

Previous research has shown that auxetic structures could provide negative Poisson’s ratio and larger self-contact area ⁴². Therefore, in this work, we have decided to use an auxetic structure to enlarge the linear sensing range and enhance the sensitivity of our sensor in terms of the contact sensing. The top row in Fig. 3(a) shows the 3D models of the sensors and the bottom row shows the corresponding physical prototypes used for structure optimization. The thickness and re-entrant angle of the auxetic structure (see Fig. 3(b)) is optimized to achieve the best nominal sensitivity and the simulation results are also validated against experiment data. Detailed description of the structure optimization is presented in ‘Method’.

The sensor with re-entrant angle of 65° and the H/L ratio of 1.60 achieved the largest nominal sensitivity according to the simulation results. Figure. 3(b) shows the optimization results and the associated experimental validation. The transparent mesh illustrates the real sensor characterization results while the solid mesh represents the simulation results. Both the experimental and simulation results suggest that the sensitivity and linear sensing range are significantly enhanced by the integrated inter-locked and auxetic structure. Twenty physical tactile sensors with the re-entrant angle of 45°, 60°, 75°, 90° and five H/L ratios were fabricated to validate the FE simulation results. The enlarged diagram of the fabricated sensor with the re-entrant angle of 45^o is presented in Figure. 3(c) The simulation output of compression tests for the sensors with different auxetic features are presented Figure. 3(d), 3(e) and Movie. 2 in the supplementary material.

Characterization of the sensor for pressure sensing

Figure. 4 illustrates the characterization of the proposed tactile sensor including the sensitivity, linear sensing range (Figs. 4(a), (b) and (c)). The sensitivity (under 10KPa of pressure, which is the threshold pressure of human touch applied during routine activities ⁴³) of the planar sensor was below 0.03 KPa^-1 but is significantly increased to 0.63 KPa^-1 by integrating the interlocked papillae and optimized auxetic structure. Meanwhile, by using the proposed structure the linear sensing range also increased from 0.1 to 0.26 MPa, which is higher compared to the values displayed by most of the published high-performance sensors ^{4,13,26,37−40}. The inter-locked features not only provide the sensor with the ability of differentiating directions of the external stimuli but also improve its sensitivity. Importantly, the integration of the auxetic structure allows for enhanced sensitivity and extends the linear sensing range. The tactile sensor is mounted onto a customized horizontal testing platform (see Supplementary Fig. 4) with a push-pull dynamometer to generate the shear force (from 0 to 5 N) while a universal testing machine is used to provide the normal compression. The shear sensitivity is quantified under the normal contact pressures ranging from 0.2 KPa to 10 KPa. The maximum shear sensitivity of 0.92 N^-1 is found under the normal pressure of 0.2 KPa and seems to be reduced with the increased contact pressure induced by the normal compression. The shear sensitivities are above 0.39 N^-1 under compression pressures ranging from 0.4 to 0.8 KPa and is reduced to 0.24 N^-1 when the normal pressure is above 1 KPa. It can be found from Fig. 4(c) that the shear sensitivity is significantly improved due to the inter-locked and auxetic structure of the sensor. The sensitivity of the planar dielectric layer is found to be 0.22 N^-1 but with the integration of the biomimetic and auxetic features this value increased to 0.92 N^-1. The lower detection limit is approximately 50 Pa as shown in Fig. 4(d), and the response time is approximately 10 ms as recorded by the multimeter. In addition, the tactile sensor is tested under repeated pressing and releasing cycles in a wide range of compressive deformation. The stable signal response is achieved after 1500 loading-unloading cycles with the pressure of 100 KPa, as presented in Fig. 4 (e). A good durability is observed based on the signal response of the last 50 cycles of compression. The sensing performance of the published sensors ^13,37−40 and the sensor introduced in this study are compared in Fig. 4(g). The effects of the biomimetic inter-locked and auxetic structure on sensor performance are presented in Fig. 4(h).

Monitoring magnitude and direction of the bone-on-bone load through the senor

To demonstrate the ability of the sensor array in detecting both the magnitude and direction of external stimuli, the sensor is directly printed onto the plateau of a human vertebra model and used to monitor the bone-on-bone contact. The physical models of the L2 and L3 vertebrae together with the inter-vertebra disc (see Figs. 5(a) to 5(c)) are 3D printed with SLA and used as the working surfaces of the proposed sensor. The corresponding negative surfaces for printing the upper layer of the sensor are also created with SLA. The printed sensor is wired with a customized electric circuit for collecting the analog outputs from the sensing elements. These are then converted into digital signals and visualized on the GUI interfaces as shown in Fig. 5(d). In this scenario, motions of the spine including lateral bending, axial rotation and flexion/extension were performed under different compression forces ranging from 25 to 100 N. In the GUI interface, the magnitude of the pressure generated by the corresponding contact force acting on each tactile element is indicated in grey color, with darker shades corresponding to higher values. The pressure tends to be intense at the side of the crushing during flexion and lateral bending. In our case, the average contact pressure increased from approximately 17 KPa to 65 KPa under the compressive forces ranging from 25 to 100N. The maximum contact pressure of 70 KPa is observed at the center of the lower vertebra a consequence of the concentrated external load. The results demonstrated that the sensor printed directly onto the vertebra performed well as expected under the arbitrary contact between the intervertebral disc and the vertebra.

The proposed tactile sensor is also printed onto the model of a proximal femoral bone to monitor the bone-on-bone load of the human hip joint (see Figs. 5(e) and (f)). A conformal printing path around the femoral head is initially defined to print the sensor. The proximal femoral head and part of the pelvis are printed to perform the motions of the hip joint during the swing of the lower lime. The test results are illustrated in the GUI interface and show a reasonable pressure distribution during the whole swing phase. Therefore, the experimental results reported above suggest that this 3D printed tactile sensor can be rapidly fabricated onto geometrically complex surfaces such as the vertebra or femoral head and used for monitoring the bone-on-bone contact. The printed sensors working on hip joint and between the lumber vertebrae are shown in Movie. 3.

Sensorimotor control of the biomimetic hand through the tactile sensor

The dimensions and shape of the tactile sensor could be customized according to the sensing area (ranging from millimeter to centimeter scale) and working surfaces, respectively. To demonstrate the scalability and adaptability of the sensor, the tactile sensor with approximate dimensions of 18🞩10 mm² is printed onto the distal phalange of the index fingers in a biomimetic anthropomorphic robotic hand as shown in Figure. 6(a). The tendon-driven biomimetic hand employed in this study was reconstructed from a male subject, including the intact bone skeleton, tendon, interphalangeal ligaments and skin. Electric motors are used to actuate the anthropomorphic robotic hand and mimic the human like kinematics and grasping quality. The distal phalange of the index is disassembled from the biomimetic hand to allow for printing the tactile sensor. A customized electric circuit is integrated to collect the analog outputs from the tactile sensing elements. A Python-based controlling algorithm is developed to convert sensor analog outputs to digital signals, providing the robotic hand with tactile feedback for grasping performance. The detailed information about the hardware and control of the robotic hand are presented in ‘Methods’ section.

Using the above set up it was possible to create a realistic scenario where the grasping action of the biomimetic hand is disturbed by an external force and the response of the tactile sensor is measured in terms of feedback. For that purpose a pulling force of 0.75 N is directly applied on the ball which counteracts the grasping movement of the hand. Figures. 6(b) and (c) show the reaction control of the biomimetic hand with and without the tactile feedback, respectively. The experimental results show that the tactile sensor is able to perceive the slippage of the held object (i.e. ball) and initiate the sensorimotor control to maintain the stable grasp. This effect is not visible in the absence of tactile feedback and the external disturbance could easily break the grasping stability.

From the above applications of the proposed tactile sensors in bone-on-bone load detection and in robotic finger tactile sensing, it can been seen that the proposed tactile sensor can be rapidly integrated, i.e. rapidly printed, onto a simulated biological system for monitoring the magnitude and direction of stimuli/load/force, and providing tactile feedback for sensorimotor control.

Characterization of the sensor for temperature sensing and the application on robotics

The temperature sensing capability is assessed by placing the sensor inside a heating chamber (see Supplementary Fig. 5 ) developed in house. The temperature variation and the signal response of the sensor are recorded to compute the temperature coefficient of resistance (TCR). TCR is a key parameter for evaluating the sensing performance of the temperature sensor and is defined as:

S= (ΔR/R₀)/ΔT (3)

where ΔR is the change of the resistance and R0 is the resistance measured at room temperature. ∆T is the change of the applied temperature.

The weight ratio of graphene to CNT is optimized to achieve the most stable and highest TCR. The detailed experimental setup and material optimization are presented in the ‘Method’ section. Figure. 7(a) shows the temperature response of the sensor under 5 thermal cycles. A stable and linear relationship between the relative resistive change and temperature variation with a sensitivity of 0.27% °C^− 1 is achieved. The temperature sensor has a sensing range from 40 to 90°C. The response of the sensor under specific temperatures of 50 and 90°C is presented in Fig. 7 (b). The devices are heated from room temperature to 50 or 90°C and maintain at these peak temperatures for approximately 1 min before cooling down to the room temperature. The sensor can respond to the temperature variation with a reasonable response time and the variation of the relative resistance change follows well with the heating profile. Figure 7(c) shows that the tactile sensor printed on the fingertip of the biomimetic hand could sense the temperature variation of a glass with hot water and provide feedback to release the firm grasping. All these characterization results revealed a reasonable temperature sensing performance of the tactile sensor and its application on the robotic hand.

Most of the tactile sensors developed over the past decade were manufactured using sophisticated technological processes such as molding ^{12,16,26,44,45}, photolithography, and etching ^17–20. Although well established, these processes normally require highly skilled labor for the operation of the systems and preparation of materials, especially when advanced composite materials are used ^11,14,16. Also, previous research work has failed to properly address the need for compliance between the tactile sensor and the working surfaces, resulting in planar sensors with poor fitting to arbitrary surfaces ^46,47. The tactile sensor introduced in this study can provide enhanced sensing capacity to detect tactile information including direction and magnitude of the stimuli, and can be rapidly fabricated in situ to fit onto the specific working surface at feasible and affordable cost. At the meantime, this tactile sensor can also response to temperature variation with a reasonable sensitivity. As presented in Fig. 4(g), the linear sensing range of our sensor is larger than the pressure sensors reported in the literature ^13,37−40. The other key sensing indicators including the normal, shear sensitivity and lower detection limit are also superior or at least comparable with those of previously reported sensors. The temperature sensing performance including the sensing range and sensitivity are also comparable with the published carbon-polymer composite based ^48–51 and commercial temperature sensors ⁵². Different from previous works, we have employed 3D printing technology to fabricate our sensor, resulting in shorter lead-times and lower production costs compared with the published high-performance tactile sensors ^{2,15,17–20,32}. A more detailed comparison is provided in Table 1 of the supplementary material.

The inter-locked structure is employed in the sensor design to mimic the biomechanical characteristics of the human skin, making the structure more sensitive to the variations of strain components and enabling the sensor to detect the direction of the stimuli. Meanwhile, the auxetic structure is optimized and validated against the experimental results. The re-entrant angle and thickness of the sensor are the main geometrical factors dominating the negative Poisson’s ratio effect of the auxetic structure. These two factors are optimized and the results suggested that the re-entrant angle of 65° and the H/L ratio of 1.60 can provide the best piezoresistive effect. The self-contact area is significantly increased by integrating the auxetic structure, leading to a wide linear sensing range. Simulation results also suggest that the negative Poisson's ratio enabled a large self-contact area which is linearly increased with the applied forces in the sensor with biomimetic and auxetic features. The sensing range is significantly increased from 0.1 to 0.26 MPa. Since the piezoresistive characteristics of the elastic polymer-based sensor is dominated by the stress/strain related parameters during external stimuli, the higher sensitivity might have been caused by the sensitive SED variation to the external stimuli according to the simulation results. The normal sensitivity was enhanced from 0.03 KPa^-1 to 0.63 KPa^-1 with the interlock and auxetic features, making the structure more responsive to the variation of normal pressure, and leading to a better piezoresistive performance. The shear sensitivity is enhanced (from 0.41 N^-1 to 0.92 N^-1 only with inter-lock features) with the optimized authentic structure whilst the lowest detection limit can reach to 50 Pa.

The sensitivity in terms of temperature sensing is not ideal (absolute value of TCR is below 0.0001°C^-1) based on the CNT/silicone rubber composite (see Supplementary, Fig. 6 (a)). Most of the twisted CNT are stretched apart due to the thermodynamic expansion of silicone rubber under the increasing temperature, leading to more conductive pathways and the reduced resistance. Therefore, CNT/silicone rubber composite shows a negative TCR. However, the stretched CNT cannot return to the original entangled status after cooling resulting in a different distribution of conductive pathways and the decreased absolute value of TCR. Therefore, the 2D graphene nano palates are added together with CNT into the silicon rubber to enhance the thermoresistive stability and sensitivity. Due to the finely distributed nano palate structure, relatively stable amount of the conductive pathways are achieved after a heating and cooling thermal cycle. The graphene palate spread out with the increased temperature which dominated the decreased conductive pathways resulting from the thermal expansion of silicone matrix. The experimental results also suggest that the temperature sensing performance including the sensitivity and stability can be enhanced by adjusting the weight ratio between two carbon nano materials (see Supplementary, Figs. 6 (b) to (d)).

The sensor in this study has shown that a better performance can be achieved by integrating the auxetic or other meta structures in the design. The application of the optimized auxetic structure may provide new insights towards the design of tactile sensors in the future. The tactile sensor can be fully printed directly and efficiently onto any uneven surface including the human phalange, proximal femur bones and the plateau of lumber vertebra perfectly compliant with the working surfaces. The experimental and simulation results suggested that benefiting from directly printing the tactile sensors onto the curved surfaces, the deformation effects on sensor sensitivity and linear sensing range were negligible, while for the other published tactile sensors, the deformation effect is still a critical issue that remains unsolved. The size and shape of the tactile sensor can be easily adjusted according to the demand of the sensing area (from mm² to cm²).

In summary, a fully 3D printed flexible tactile sensor based on biomimetic inter-lock and auxetic structure is developed for sensing the contact pressures and the environmental temperature. Using a customized 3D printer a Graphene/CNT/Silicone rubber and silver-coated powder-silicone composite was printed directly onto the working surfaces, fabricating the tactile sensors efficiently and economically. Similar performance to most of the published high-performance tactile sensors is achieved ^13,37−40. The integrated biomimetic and auxetic structure dominates the high sensitivity and large linear sensing range in terms of the pressure/force sensing, the optimized weight ration of CNT to Graphene of the composite material ensured the good temperature sensing performance. The auxetic structure proposed in this study can be effectively used in the structural design of tactile sensors due to its unique mechanical properties. The sensors are rapidly fabricated onto the working surfaces such as the distal phalangeal bone, human vertebra and distal femur bone to monitor the sophisticated biomechanical contact. Sensorimotor performance was achieved by the robotic hand based on the tactile and temperature feedback from the proposed tactile sensor. Future work will encompass the use of this sensor on tactile sensing actions of robotic hand, including realizing light contact with an unknown held object under sensorimotor control which is a critical aspect to human-robot interaction and prosthetics development ⁵³.

Material preparation for 3D printing

The high purity (95 wt%) Multi-Walled CNT (XFM25, XFNano, China) and graphene nano palate (XF021, XFNano, China) are dispersed uniformly in isopropanol by ultrasonic dispersion at room temperature for 30 min to obtain a good CNT /Graphene suspension (see Fig. 2(a)). The silicone main agent (Polycraft GP3481-F Silicone rubber, MB Fibreglass, UK) is added into the CNT/Graphene isopropanol suspension and heated to 80 ℃ under mechanical stirring until the isopropanol is completely evaporated. After cooling down to the room temperature, the silicone curing agent is added to the composite in a weight ratio of 1:12. The graphene palate are dispersed evenly in silicon matrix and no obvious aggregation of CNT (see Supplementary, Fig. 7) is found based on micrographs. Our experimental results suggest that conductivity (see Supplementary, Fig. 8) and piezoresistive effect of the tactile sensor in terms of pressure sensing can be achieved based on the CNT/Silicon rubber composite with 7.5 wt% of the CNT. However, the magnitude of TCR based on the CNT/Silicon rubber composite is below 0.1%℃^-1 and not stable after each heating cycle (see Supplementary, Fig. 6 (a)). To enhance the thermoresistive effect of the sensor, graphene nano palate is added and the weight ratio of CNT to graphene is optimized to achieve the best temperature sensing performance while maintaining a comparable piezoresistive effect with the CNT/Silicon rubber composite. The conductivity of the CNT/ Silicon rubber composite is superior to those of Graphene/CNT/Silicon rubber composite. To avoid a significant effect on the piezoresistive effect and maintain a stable extrusion quality during printing, the total weight percentage of the carbon-based nano material is set to be constant at 7.5% while the content of graphene palate is maintained below 4%. The thermal-resistive performance can only be achieve when the content of graphene palate is above 3.5% wt. Therefore, the tactile sensors with weight ratios of CNT to graphene of 4:3.5, 3.75:3.75 and 3.5:4 are fabricated and the temperature sensing performances are characterized (see Supplementary, Figs. 6(b) to (d)). The CNT/Graphene/ Silicon rubber composite with 3.5 wt% of Multi-Walled CNT and 4 wt% Graphene is found can achieve the good stability and the highest TCR value among the sensors with different weight ratios of two carbon nano materials.

Silver coated copper-powder-silicone composite (see Fig. 2(a)) was used to print the electrode of the sensor. It’s prepared by mixing the silver coated copper palate (48 µm) with silicone rubber with the weight ratio of 1:3.5. The coupling agent (KH550) is then added under the weight ratio of 1:100 and mechanically mixed with the composite to improve the conductivity.

3D printing platform and optimization of the printing configurations

A customized 3D printing platform was employed on the fabrication of the sensors. The system includes a 3-axis motion gantry with a workspace of 250 × 250 × 150 mm³, linear motion converters (THK, Japan), two-stage gear reducers (Oriental Motor Ltd., Japan), and servo motors (MiGe, China). A high precision dispenser (S-SIGMA-X3-V5, Musashi Engineering, Japan) is integrated with a flexible Teflon tube with a diameter of 5.5 mm and a syringe as the material feeding device. The control system contains a 6-axis controller and self-customized software. The 6-axis controller (Leadshine Technology Ltd., China) is used to process the G code from the software and control the motions of the feeding syringes and the 3-axis stage. This customized 3D printing platform is employed to execute non-linearly varying material printing in our previous research [32] and high printing resolution and quality are achieved.

To achieve the stable extrusion performance, the printing parameters including the printing speed and air pressure are optimized. The moving speed ranging from 5 to 15 mm/s (step increment 1mm/s) under the pressure varied from 0.3 to 0.7 MPa (step increment 0.1 MPa) were used to carry out the trail line printing, resulting in totally 121 different combinations of printing parameters. The pressure of 0.4 MPa with a printing speed of 12 mm/s was found to achieve a stable printing quality and appropriate extrusion width.

During the 3D printing process, in Step 1, the lower electrode layer is printed with a 15-gauge nozzle mounted on the syringe loaded with silver coated copper-silicone composite material. Then, another syringe with the Graphene/CNT/Silicon rubber composite is used and mounted onto a 22-gauge nozzle for line printing the auxetic structure of the lower dielectric layer. The dot printing is controlled manually to print the large papilla structure of the lower layer. The same printing parameters and patterns are applied to print the upper dielectric layer and then attached with the lower layer. Finally, the upper electrode is printed on top of the sensor similar to that of the lower electrode in Step 1. The 3D printing process of the sensor is shown in Movie 1 in the supplementally materials.

Testing platform for characterization of pressure sensing

In order to take full advantage of the piezoresistive and mechanical properties of the interlocked-auxetic features, the standard sensor with the sensing area of 18🞩20 mm² is fabricated for the characterization of pressure sensing (see Fig. 4). The sensor prototype is powered by 5V DC supply and connected into a customized electronic circuit. The output current is measured and collected by the multimeter (Keysight 34465A, Keysight Ltd., HK) at the frequency of 500 Hz. The normal pressure applied onto the sensor is controlled by the universal testing machine (WH-5000, Weiheng Co., China) (above 10KPa) and a precise push-pull dynamometer (below 10KPa). The push-pull dynamometer is mounted onto a horizontal tensile test platform (see Supplementary, Fig. 4) to produce shear force for charactering piezoresistive effects under the stimuli from horizontal directions.

Testing platform for characterizations of temperature sensing

A in-house heating chamber composed a plastic box with the dimension of 75🞩30🞩15 mm³ and a heating plate powered by DC voltage supply is developed to provide a stable temperature variation. The variation of the resistance is recorded by multimeter (Keysight 34465A, Keysight Ltd., HK) while the temperature is monitored with another multimeter (see Supplementary, Fig. 5).

The FE modelling and optimization of the sensor structure

The tactile sensors with different auxetic structures are designed and modelled in Creo Parametric (PTC Inc. Boston, US). Two key geometric parameters including the re-entrant angle (from 45° to 90° with the increment of 1°) and the H/L (Fig. 3(b)) ratio (H/L = 2.7, 2.0, 1.6 1.3 1.0) are optimized. 230 tactile sensors with different gematrical parameters are modelled for structural optimization (46 re-entrant angles together with 5 H/L ratio, resulting in 46 🞩 5 = 230 different auxetic structures). The compression test is simulated in commercial FE software Abaqus. The heights of the inter-locked large and small papilla structures are adjusted with the auxetic structure as\(L\times sin\theta\) while the diameters remain constant. The sensitivity in terms of the piezoresistive effect is defined as the objective of the structure optimization for this tactile sensor in the linear regime as:

S = (ΔI/I₀)/ΔP (1)

where ∆I is the change of current flow over the dielectric layers, I₀ is the base current measured without any external stimuli on the sensor, and ∆P is the change of the applied contact pressure.

Further, the nominal sensitivity, denoted as Sn, for optimizing the auxetic structure was defined as:

Sn = (ΔSED/SED₀)/ΔP (2)

where ∆SED stands for the change of the strain energy density at the site of the sensing elements, and SED₀ is the strain energy density computed under the maximum pressure of 0.26 MPa.

Research on piezoresistive characteristics of the elastic polymer-based sensor has shown that the sensitivity is dominated by the geometry of the structure and the stress/strain variation within the composite ^54–56. Due to the relatively small variations in the geometry of thin film-like flexible sensors, the change of strain energy density relative to the external contact force is considered as the main factor affecting the piezoresistive effect. Therefore, the variation of SED is employed instead of current to evaluate the nominal sensitivity for the structure optimization and simulation.

The material properties for the sensors are defined as linear elastic with the Young’s Modulus of 3.0 MPa and Poisson’s ratio of 0.3. A flat plate is modelled to compress the sensor to a specified displacement. The bottom surface of the sensor is fixed and the mesh size of 0.5 mm with tetrahedral element is used for the simulation. ‘Hard contact’ is assigned between the flat plate and the sensor, the ‘self-contact’ is defined over all the external and internal surfaces of the sensor to measure the self-contact area during the compression. The strain energy density is extracted from all the elements for deriving the nominal sensor sensitivity. In total 230 simulations with different auxetic structures are carried out for the optimization to find the largest nominal sensitivity under the same external stimuli. The simulation results are also validated against the experimental data, a good agreement is achieved between nominal and real sensitivity. Therefore, the FE simulation provided reliable optimization results of the structural design.

The bone-on-bone load monitoring and sensorimotor control of the biomimetic hand with the tactile sensor

The sensor is printed onto the lumber vertebra and the distal femur head, connected with a customized circuit composed of two shift registers, two multiplexers and an Arduino Uno^® board for collecting the analog output from the tactile sensing elements. The analog output is converted into digital signal and visualized through the GUI programmed in Processing IDE (Processing.org).

A tendon-driven biomimetic hand containing intact hand bone skeleton, interphalangeal ligaments, tendon and skin is employed in this study to demonstrate the sensorimotor control with the 3D printed tactile sensor. The skeleton of the hand is 3D printed with polylactic acid (PLA) and the soft-tissues are modeled with silicone-rubber. The anthropomorphic size of the biomimetic hand is reconstructed based on the human hand from a 23-year-old male subject. The sensor is 3D printed onto the distal phalange of the index finger, connected with the same customized electric circuit to collect the analog output from the tactile sensing elements and convert it to digital signals as the feedback for controlling the biomimetic hand. Fishing lines are connected with the bone skeleton and driven by the electric motors (Dynamixel MX-12W, Robotics Inc., US) imitating the tendon-driven mechanism of human hand. The pulling force along the fishing line provided by the electric motors is scaled with muscle contraction forces of the human subject through a dynamometer. The in-vivo grasping experiment is carried out and the muscle forces under the reactive touch are derived based on the electromyographic signals collected through the Delsys system (Delsys Trigno, Delsys Inc., US). The subject gave informed consent to participate in the grasping experiments, which were approved by the Ethics Committee of the First Hospital of Jilin University. A python program is developed for processing the pressure signal and controlling the motors to produce the pulling forces with the preset magnitudes similar with the muscle contraction forces of human subject. The human-like sensorimotor performance is restored on this biomimetic hand based on the tactile feedback, ensuring its application on the next-generation neuro-prosthetic.

Contributions

YW and LR conceived the research idea. YW designed and fabricated the sensor, performed the experiments, analysed the data and wrote the paper. MD provided scientific guidance and revised the paper. BL, YZ and LY performed some experiments. LR and GW supervised the research work, analysed the data and revised the manuscript.

Data Availability

The source data including Supplementary Figures 1-8, Table 1 and Supplementary Movie 1-4 are provided with this paper. The 3D printing process of the tactile sensor on a finger bone is presented in Movie 1. Movie 2 shows the FE simulations for the optimization of the auxetic structure of the sensor. Movie 3 displays the measurement of the intervertebral disc pressure and the mechanical load acting at the hip joint using the 3D printed sensor. Movie 4 presents the measurement of the finger contact pressure and sensorimotor control of a biomimetic hand based on the tactile feedback of the 3D printed sensor. The 3D model and 2D drawings of the proposed tactile sensor can be found on figshare (https://figshare.com/articles/figure/3D_Model_and_Drawings_of_the_Tactile_Sensor_NatCom_2021/16569696). Data supporting the findings of this manuscript are available from the corresponding author upon reasonable request.

Code Availability

The G code for 3D printing the sensor and the Python code for performing sensorimotor control on biomimetic hand are available on OSF (https://osf.io/p2uyt/?view_only=d62014436e414ce69ab22f0a496033b3).

Acknowledgements

We would like to thank our research group members in University of Manchester and Jilin University for their great support and assistant to this study.

Funding: This research was partly supported by the project of National Key R&D Program of China (No. 2018YFC2001300), the project of National Natural Science Foundation of China (No. 91948302, No. 91848204, No. 52005209, and No. 51675222).

Competing interests: The authors declare that they have no competing interests.

Hasan, S. A. U. et al. A sensitivity enhanced MWCNT/PDMS tactile sensor using micropillars and low energy Ar+ ion beam treatment. Sensors 16, 93 (2016).
Boutry, C. M. et al. A hierarchically patterned, bioinspired e-skin able to detect the direction of applied pressure for robotics. Science Robotics 3 (2018).
Miao, P. et al. Graphene nanostructure-based tactile sensors for electronic skin applications. Nano-Micro Letters 11, 1-37 (2019).
Li, X. et al. Highly sensitive flexible tactile sensors based on microstructured multiwall carbon nanotube arrays. Scripta Materialia 129, 61-64 (2017).
Park, J. et al. Tactile-direction-sensitive and stretchable electronic skins based on human-skin-inspired interlocked microstructures. ACS nano 8, 12020-12029 (2014).
Qin, Z. et al. Carbon nanotubes/hydrophobically associated hydrogels as ultrastretchable, highly sensitive, stable strain, and pressure sensors. ACS applied materials & interfaces 12, 4944-4953 (2020).
Chen, X. et al. Flexible three-axial tactile sensors with microstructure-enhanced piezoelectric effect and specially-arranged piezoelectric arrays. Smart Materials and Structures 27, 025018 (2018).
Choi, E. et al. Highly sensitive tactile shear sensor using spatially digitized contact electrodes. Sensors 19, 1300 (2019).
Luo, N. et al. Hollow‐structured graphene–silicone‐composite‐based piezoresistive sensors: Decoupled property tuning and bending reliability. Advanced materials 29, 1702675 (2017).
Wang, Z. et al. Full 3D printing of stretchable piezoresistive sensor with hierarchical porosity and multimodulus architecture. Advanced Functional Materials 29, 1807569 (2019).
Ha, M. et al. Bioinspired interlocked and hierarchical design of ZnO nanowire arrays for static and dynamic pressure‐sensitive electronic skins. Advanced Functional Materials 25, 2841-2849 (2015).
Jung, M. et al. Transparent and flexible Mayan-pyramid-based pressure sensor using facile-transferred indium tin oxide for bimodal sensor applications. Scientific reports 9, 1-11 (2019).
Kwon, D. et al. Highly sensitive, flexible, and wearable pressure sensor based on a giant piezocapacitive effect of three-dimensional microporous elastomeric dielectric layer. ACS applied materials & interfaces 8, 16922-16931 (2016).
Lee, Y. et al. Flexible ferroelectric sensors with ultrahigh pressure sensitivity and linear response over exceptionally broad pressure range. ACS nano 12, 4045-4054 (2018).
Park, J., Kim, M., Lee, Y., Lee, H. S. & Ko, H. Fingertip skin–inspired microstructured ferroelectric skins discriminate static/dynamic pressure and temperature stimuli. Science advances 1, e1500661 (2015).
Wang, Y., Zhu, L., Mei, D. & Zhu, W. A highly flexible tactile sensor with an interlocked truncated sawtooth structure based on stretchable graphene/silver/silicone rubber composites. Journal of Materials Chemistry C 7, 8669-8679 (2019).
Makihata, M. et al. Design and fabrication technology of low profile tactile sensor with digital interface for whole body robot skin. Sensors 18, 2374 (2018).
Oh, H., Yi, G.-C., Yip, M. & Dayeh, S. A. Scalable tactile sensor arrays on flexible substrates with high spatiotemporal resolution enabling slip and grip for closed-loop robotics. Science advances 6, eabd7795 (2020).
Pyo, S., Lee, J.-I., Kim, M.-O., Lee, H.-K. & Kim, J. Polymer-based flexible and multi-directional tactile sensor with multiple NiCr piezoresistors. Micro and Nano Systems Letters 7, 1-8 (2019).
Zhu, B. et al. Microstructured graphene arrays for highly sensitive flexible tactile sensors. Small 10, 3625-3631 (2014).
Büscher, G. H., Koiva, R., Schürmann, C., Haschke, R. & Ritter, H. J. Flexible and stretchable fabric-based tactile sensor. Robotics and Autonomous Systems 63, 244-252 (2015).
Abbass, Y., Saleh, M., Dosen, S. & Valle, M. Embedded Electrotactile Feedback System for Hand Prostheses Using Matrix Electrode and Electronic Skin. IEEE Transactions on Biomedical Circuits and Systems 15, 912-925 (2021).
Tomo, T. P. et al. A new silicone structure for uSkin—A soft, distributed, digital 3-axis skin sensor and its integration on the humanoid robot iCub. IEEE Robotics and Automation Letters 3, 2584-2591 (2018).
Yuan, W., Dong, S. & Adelson, E. H. Gelsight: High-resolution robot tactile sensors for estimating geometry and force. Sensors 17, 2762 (2017).
Ward-Cherrier, B. et al. The tactip family: Soft optical tactile sensors with 3d-printed biomimetic morphologies. Soft robotics 5, 216-227 (2018).
Sun, X. et al. Flexible tactile electronic skin sensor with 3D force detection based on porous CNTs/PDMS nanocomposites. Nano-Micro Letters 11, 57 (2019).
Kou, H. et al. Wireless wide-range pressure sensor based on graphene/PDMS sponge for tactile monitoring. Scientific reports 9, 1-7 (2019).
Camilli, L. & Passacantando, M. Advances on sensors based on carbon nanotubes. Chemosensors 6, 62 (2018).
Ma, P.-C. et al. Enhanced electrical conductivity of nanocomposites containing hybrid fillers of carbon nanotubes and carbon black. ACS applied materials & interfaces 1, 1090-1096 (2009).
Obitayo, W. & Liu, T. A review: Carbon nanotube-based piezoresistive strain sensors. Journal of Sensors 2012 (2012).
Yi, Z., Zhang, Y. & Peters, J. Biomimetic tactile sensors and signal processing with spike trains: A review. Sensors and Actuators A: Physical 269, 41-52 (2018).
Pang, C. et al. A flexible and highly sensitive strain-gauge sensor using reversible interlocking of nanofibres. Nature materials 11, 795-801 (2012).
Koç, İ. M. & Akça, E. Design of a piezoelectric based tactile sensor with bio-inspired micro/nano-pillars. Tribology International 59, 321-331 (2013).
Devaraj, H. et al. Bio-inspired flow sensor from printed PEDOT: PSS micro-hairs. Bioinspiration & biomimetics 10, 016017 (2015).
Wan, Y. et al. A Highly Sensitive Flexible Capacitive Tactile Sensor with Sparse and High‐Aspect‐Ratio Microstructures. Advanced Electronic Materials 4, 1700586 (2018).
Guo, S. Z., Qiu, K., Meng, F., Park, S. H. & McAlpine, M. C. 3D printed stretchable tactile sensors. Advanced Materials 29, 1701218 (2017).
Zhang, J. et al. Highly sensitive flexible three-axis tactile sensors based on the interface contact resistance of microstructured graphene. Nanoscale 10, 7387-7395 (2018).
Yang, L. et al. Development of a highly sensitive, broad-range hierarchically structured reduced graphene oxide/polyhipe foam for pressure sensing. ACS applied materials & interfaces 11, 4318-4327 (2019).
Wang, X., Gu, Y., Xiong, Z., Cui, Z. & Zhang, T. Silk‐molded flexible, ultrasensitive, and highly stable electronic skin for monitoring human physiological signals. Advanced materials 26, 1336-1342 (2014).
Mannsfeld, S. C. et al. Highly sensitive flexible pressure sensors with microstructured rubber dielectric layers. Nature materials 9, 859-864 (2010).
Johnson, K. O. The roles and functions of cutaneous mechanoreceptors. Current opinion in neurobiology 11, 455-461 (2001).
Evans, K. E., Nkansah, M., Hutchinson, I. & Rogers, S. Molecular network design. Nature 353, 124-124 (1991).
Hu, F. et al. Gel-Based Artificial Photonic Skin to Sense a Gentle Touch by Reflection. ACS applied materials & interfaces 11, 15195-15200 (2019).
Ji, Z. et al. The design and characterization of a flexible tactile sensing array for robot skin. Sensors 16, 2001 (2016).
Kang, K., Cho, Y. & Yu, K. J. Novel nano-materials and nano-fabrication techniques for flexible electronic systems. Micromachines 9, 263 (2018).
Tiwana, M. I., Redmond, S. J. & Lovell, N. H. A review of tactile sensing technologies with applications in biomedical engineering. Sensors and Actuators A: physical 179, 17-31 (2012).
Stassi, S., Cauda, V., Canavese, G. & Pirri, C. F. Flexible tactile sensing based on piezoresistive composites: A review. Sensors 14, 5296-5332 (2014).
Yang, J. et al. Wearable temperature sensor based on graphene nanowalls. Rsc Advances 5, 25609-25615 (2015).
Liu, G. et al. A flexible temperature sensor based on reduced graphene oxide for robot skin used in internet of things. Sensors 18, 1400 (2018).
Dan, L. & Elias, A. L. Flexible and Stretchable Temperature Sensors Fabricated Using Solution‐Processable Conductive Polymer Composites. Advanced Healthcare Materials 9, 2000380 (2020).
Wang, Z. et al. 3D-printed graphene/polydimethylsiloxane composites for stretchable and strain-insensitive temperature sensors. ACS applied materials & interfaces 11, 1344-1352 (2018).
Kuo, J. T., Yu, L. & Meng, E. Micromachined thermal flow sensors—A review. Micromachines 3, 550-573 (2012).
Lepora, N. F. et al. in 2021 IEEE International Conference on Robotics and Automation (ICRA). 1622-1628 (IEEE).
Kanda, Y. Piezoresistance effect of silicon. Sensors and Actuators A: Physical 28, 83-91 (1991).
Matsuda, K., Suzuki, K., Yamamura, K. & Kanda, Y. Nonlinear piezoresistance effects in silicon. Journal of applied physics 73, 1838-1847 (1993).
Toriyama, T., Tanimoto, Y. & Sugiyama, S. Single crystal silicon nano-wire piezoresistors for mechanical sensors. Journal of microelectromechanical systems 11, 605-611 (2002).

There is NO Competing Interest.

Download PDF

Journal Publication

published 14 Nov, 2023

Read the published version in Communications Engineering →

Version 2

posted

You are reading this latest preprint version

Fully 3D Printed Flexible, Conformal and Multi-directional Tactile Sensor with Integrated Biomimetic and Auxetic Structure

Status:

Journal Publication

Version 2

Abstract

Figures

Introduction

Results

The tactile sensor – Design and 3D printing

Auxetic structure optimization of the tactile sensor

Characterization of the sensor for pressure sensing

Monitoring magnitude and direction of the bone-on-bone load through the senor

Sensorimotor control of the biomimetic hand through the tactile sensor

Characterization of the sensor for temperature sensing and the application on robotics

S= (ΔR/R₀)/ΔT (3)

Discussion

Methods

Material preparation for 3D printing

3D printing platform and optimization of the printing configurations

Testing platform for characterization of pressure sensing

Testing platform for characterizations of temperature sensing

The FE modelling and optimization of the sensor structure

S = (ΔI/I₀)/ΔP (1)

The bone-on-bone load monitoring and sensorimotor control of the biomimetic hand with the tactile sensor

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 2

Fully 3D Printed Flexible, Conformal and Multi-directional Tactile Sensor with Integrated Biomimetic and Auxetic Structure

Status:

Journal Publication

Version 2

Abstract

Figures

Introduction

Results

The tactile sensor – Design and 3D printing

Auxetic structure optimization of the tactile sensor

Characterization of the sensor for pressure sensing

Monitoring magnitude and direction of the bone-on-bone load through the senor

Sensorimotor control of the biomimetic hand through the tactile sensor

Characterization of the sensor for temperature sensing and the application on robotics

S= (ΔR/R0)/ΔT (3)

Discussion

Methods

Material preparation for 3D printing

3D printing platform and optimization of the printing configurations

Testing platform for characterization of pressure sensing

Testing platform for characterizations of temperature sensing

The FE modelling and optimization of the sensor structure

S = (ΔI/I0)/ΔP (1)

The bone-on-bone load monitoring and sensorimotor control of the biomimetic hand with the tactile sensor

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 2

S= (ΔR/R₀)/ΔT (3)

S = (ΔI/I₀)/ΔP (1)