Integrated Simultaneous Detection of Tactile and Bending Cues for Soft Robotics

Abstract

Soft robots should move in an unstructured environment and explore it and, to do so, they should be able to measure and distinguish proprioceptive and exteroceptive stimuli. This can be done by embedding mechanosensing systems in the body of the robot. Here, we present a polydimethylsiloxane block sensorized with an electro-optical system and a resistive strain gauge made with the supersonic cluster beam implantation (SCBI) technique. We show how to integrate these sensing elements during the whole fabrication process of the soft body and we demonstrate that their presence does not change the mechanical properties of the bulk material. Exploiting the position of both sensing systems and a proper combination of the output signals, we present a strategy to measure simultaneously external pressure and positive/negative bending of the body. In particular, the optical system can reveal any mechanical stimulation (external from the soft block or due to its own deformation), while the resistive strain gauge is insensitive to the external pressure, but sensitive to the bending of the body. This solution, here applied to a simple block of soft material, could be extended to the whole body of a soft robot. This approach provides detection and discrimination of the two stimuli (pressure and bending), with low computational effort and without significant mechanical constraint.

Introduction

In recent years, the field of soft robotics^1–7 is introducing many new concepts for designing robotic systems, which today work in a well-defined environment and perform precise tasks. The strive for developing soft robots is fueled by the aspiration of building systems, made of nonrigid materials, that move and operate in an unstructured environment,^8–11 and interacting more naturally and safely with humans. Hence, strong challenges are posed in areas of modeling, design, fabrication, and control. In this context, we focus on new strategies that will enable future soft robots of both exteroception and proprioception capabilities. Our aim is to create new technological solutions for soft mechanosensing, to allow soft robots perceiving external objects by touch, as well as reconstructing their shape and deformation.^12,13

Materials used for building soft robots are mainly stretchable elastomers, such as platinum catalyzed silicones (i.e., polydimethylsiloxane [PDMS] or Ecoflex^® series). While elastic for small deformations, these materials exhibit a highly nonlinear behavior for large strains.¹² At the same time, their mechanical characteristics are useful to attempt mimicking natural skin and underlying tissues and they are a good choice for the fabrication of bioinspired robots. To our knowledge, only Shepherd's group has uniquely embedded proprioception and exteroception in a soft robot inspired by the octopus,¹⁴ by way of a highly stretchable skin. Owing to the capacitive and electroluminescent sensing functions of the skin, it was demonstrated that the robot can either sense the skin stretching, due to its own movement, or that it can detect a pressure coming from the outside.

In the last decade, soft mechanical sensing solutions have shown a strong boost.^15–19 This has mainly been pushed by the exploitation of novel materials (e.g., graphene, metallic nanowires, and functional organic compounds) and to the rapid development of fabrication techniques at the micro/nanoscale. In particular, highly deformable materials (both conductive and nonconductive) were used to mimic the basic characteristics of skin-like flexibility or elasticity. For instance, strain or bending sensors, based on silver nanowires,^20,21 graphene,^22–24 or carbon nanotubes (CNT),^25,26 have been embedded in ultrathin polymeric films. Also, sensors based on gold nanowires,²⁷ organic transistors,^28–30 and conductive textiles have been proposed.^31–33 In addition to single sensors, electronic skins (e-skin) have shown remarkable advancements, especially in the field of ultrathin and ultraconformable systems.^14,34–40

However, in soft robots and wearable systems,^41–46 it becomes crucial using such deformable devices to encode a variety of mechanical inputs. In doing so, several aspects are still overlooked, owing to testing single specific functionalities of the devices or to implementing them per se. One key issue is the capability of detecting and distinguishing different types of mechanical cues either external or originated by the movement of the artifacts (e.g., pressure, strain, or bending), which we name mechanical multimodality.

One possible strategy should be the integration of separate transducers, each sensing one different mechanical stimulus. Nevertheless, in general, soft tactile sensing devices respond in a very similar way under different solicitations. This is mainly due to their basic building blocks made of soft materials that exhibit a very similar response (e.g., mostly electrical, such as change in resistance, capacitance, or current) to deformations induced in the device by any type of outer mechanical stimuli (e.g., indentation and traction). Most of the multimodal mechanical sensors found in the literature suffer from such limitation.^47–49 Another possible strategy is the integration of many similar sensors, each transducing the same mechanical solicitation and elaborating the collected sensor signals with specifically developed algorithms to discriminate between different stimulations.

Following this approach, in a previous work, we addressed the discrimination of two different types of mechanical stimuli in a soft body.⁵⁰ We used two similar sensing elements (based on e-textiles and elastomers embedded in the soft module) to recognize the convex from the concave side and detect the bending angle (or maximum deflection), and to distinguish between bending (i.e., local strain) and external force solicitation. However, when this solution is distributed along the body of the whole robot, it can be limited by the high number of connections, which would affect the mechanical behavior of the system, and by the electronic and computational complexities. For example, in Cianchetti et al.,⁵¹ resistive stretch sensors are exploited to reconstruct the spatial configuration of a continuum soft arm. In this case, the local curvature of the arm is retrieved with differential strain measurements, and the reconstruction method allows obtaining the spatial configuration from local deformations. Yet, the bottleneck for real applications is the wiring needed for measuring each resistance variation. Another example is given by Chossat et al.,⁵² who developed a soft tactile skin using an embedded ionic liquid matrix. Here, a tomographic imaging process is needed for reconstructing the stimulation map, and for larger matrices, real-time elaboration is computationally very demanding.

Indeed, another and very promising strategy is to combine material structuring at the micro/nanoscale with nanocomposites, such as graphene, nanowires, or CNT. For instance, following this approach a multilayer device, made of a porous rubber and an air gap between two PDMS/CNT conductive thin films, was developed to distinguish between pressure and lateral strain.⁵³ Similarly, a device based on nylon and polyurethane fibers, functionalized with silver nanowires and coated with piezoresistive rubber,⁵⁴ has the capability to discriminate pressure, strain, and flexion.

Nevertheless, even with remarkable sensing capabilities such as the abovementioned example, the current devices are optimized as single components, without considering the direct integration in a soft body. Also, they are tested on rigid supports of the experimental setups. This represents another key issue that must be tackled for sensorizing soft robots (i.e., where the substrate is the robot body made of soft elastomers such as PDMS or Ecoflex) and wearable systems (i.e., where the substrate is the human body).

Proposed solution

In this article, we present a new solution for soft mechanosensing that has three key aspects. The first is that it can retrieve both exteroceptive (pressure) and proprioceptive (strain due to bending) information; the second is that the two transduction principles (i.e., electro-optical and resistive) are exploited in a soft body; finally, the third is that the sensing technologies used do not alter the original mechanical characteristics of the starting soft material (i.e., PDMS) intended to constitute the bulk of a future actuated structure.

More in detail, our device is made of a PDMS block (that we call body) with optical components embedded during the fabrication process. A photodiode and a phototransistor are used as transmitter and receiver, respectively, obtaining an electro-optical sensor. The use of PDMS as soft waveguide for sensing mechanical stimulations has been already reported. For instance, Ramuz et al.⁵⁵ presented a transparent, pressure-sensitive, artificial skin based on a layer of PDMS embedding organic LEDs and photodetectors, while we have previously shown how the same principle can be used to develop an extended electronic skin for distributed and multiple pressure mapping.⁵⁶ Also, very recently, a fiber-reinforced soft prosthetic hand was reported, in which PDMS-based waveguides are exploited to detect curvature, elongation, or external force.⁵⁷ All these results are encouraging and show the potential of using a soft transparent material having the twofold function of mechanical substrate and waveguide.

In this work, we use the waveguide system for sensing pressure. Moreover, one of the body surfaces embeds patterned conductive electrodes forming a resistive strain gauge, which are produced by supersonic cluster beam implantation (SCBI), an innovative technique for making conductive parts in soft elastomers.⁵⁸ The discrimination of outer pressure from bending is obtained by combining optical and resistive signals, and by using a proper architecture as discussed in the next section. In this approach, the body and the sensing elements are completely merged during the whole fabrication process. We demonstrate that all embedded sensing elements do not affect the mechanical behavior of the soft body. Finally, we show that it is possible to distinguish and to measure both sequential and simultaneous pressure/bending stimulations.

Materials and Methods

Design and strategy for revealing simultaneous multiple stimuli

PDMS is one of the most used materials in soft robotics and the device here proposed (shown in Fig. 1) can be considered as a proof-of-concept for a section of soft robot with embedded sensing capabilities. The bulk body acts as a waveguide for the light emitted by the photodiode and collected by the phototransistor. Due to the reduced dimensions of the optical elements, the bulk mechanical characteristics are not affected by their integration in the body.

FIG. 1.

Fabrication process of the device. (a) PDMS molding for making the bulk body with embedded optical elements. (b) Supersonic cluster beam process for the implantation of gold nanoparticle-based resistive strain gauge, with the implanted surface shown in the inset. (c) Soft contact connection to external wires of the implanted strain gauge by means of a nickel graphite-filled silicone adhesive. (d) Insulation and protection of the implanted surface with a thin Ecoflex^® layer deposited by spin coating. PDMS, polydimethylsiloxane.

The conductive pattern is implanted on one of the two surfaces by SCBI. A bending of the structure induces a strain on the patterned surface, which can be measured from its resistance variation. For this reason, the resistive layer cannot be located in the middle of the structure, otherwise it would be very close to the neutral plane of the device, making the strain gauge insensitive to bending (see Supplementary Data for details; Supplementary Data are available online at www.liebertpub.com/soro). This must be considered as a design guideline for a future soft robot.

In a previous work,⁵⁸ the characterization of linear strain (i.e., resistance variation vs. elongation of metal layer) was presented, demonstrating the effectiveness of this kind of implantation for the implementation of a highly reliable and sensitive strain gauge. From this resistance variation, the positive or negative bending in a 2D plane can be revealed. Also, the resistive strain gauge is almost insensitive to a pressure applied on the nonimplanted surface, since the PDMS thickness highly limits pressure transmission. Remarkably, in contrast to other available technologies used to implant or integrate conductive paths on a soft substrate, the fabrication of conductive patterns based on a metal/polymer nanocomposite produced by SCBI ensures a complete integration of the sensing elements in the bulk, with excellent mechanical stability under deformation, while the mechanical properties of the nanocomposite (Young modulus) are substantially unchanged compared to the pristine polymer.⁵⁹

In contrast, any mechanical deformation of the body can be revealed measuring the output signal of the phototransistor, which is proportional to the incident light collected by the receiver. Then, by combining this asymmetrical behavior, the system is able to distinguish clearly an outer pressure from the positive or negative bending of the body.

This sensing strategy is quite general, since it depends weakly on the soft body shape and on the materials used. A key point is to exploit different sensing principles completely decoupled from both the transduction and the mechanical point of view. However, at the same time, the sensing elements should be completely embedded in the body without interfering on the bulk mechanical properties. Regarding the bulk material, a compromise in its elasticity is needed, since a completely rigid body will not be able to reveal an applied pressure, while a very soft one will not shield the buried strain sensor. For these reasons, we chose a basic body shape (a parallelepiped) for experimental validation, and PDMS as soft body material, which is widely used in soft robots and is transparent to infrared (IR) light. Finally, we should note that the described strategy is strictly valid for movement/stimulation in a 2D plane. However, it can be the basis for reconstructing movements and stimulations in the 3D space, by combining several sensing elements.

Device fabrication

The whole fabrication process is depicted in Figure 1. First of all, the PDMS body, with dimensions of 40 × 15 × 5 mm, is obtained by curing degassed PDMS (Dow Corning; Sylgard 184), with a prepolymer/crosslinker ratio of 10:1 by weight, for 24 h at room temperature in a rigid Plexiglas^® mold. Before pouring the PDMS, an IR photodiode (VSMY1850; Vishay Semiconductors) and an IR phototransistor (TEMT7100X01; Vishay Semiconductors) are fixed into the mold, to be embedded in the PDMS block after the curing phase.

In addition, one of the body surfaces has patterned conductive electrodes, produced by an SCBI apparatus equipped with a pulsed microplasma cluster source (PMCS). PMCS consists in a ceramic body with a cavity in which a solid gold target is vaporized by a localized electrical discharge supported by a pulsed injection of an inert gas (He or Ar) at high pressure (40 bar). The metal atoms, sputtered from the target, aggregate in the source cavity to form metal clusters; the mixture of clusters and inert gas expands through a nozzle forming a supersonic beam into an expansion chamber kept at a pressure of about 10⁻⁶ mbar. Electrically neutral nanoparticles exiting the PMCS are aerodynamically accelerated in a highly collimated beam with divergence lower than 1° and with a kinetic energy of roughly 0.5 eV·atom⁻¹. The central part of the cluster beam enters, through a skimmer, a second vacuum chamber (deposition chamber) where the beam is intercepted by the polymeric substrate. Using a stencil mask and substrate rastering in the plane perpendicular to the cluster beam (Fig. 1b), it is possible to homogenously implant gold nanoparticles into the target polymer with the required pattern.

In this way, a fully stretchable and conductive pattern is obtained on the surface,⁵⁸ as shown in Figure 1b. The amount of nanoparticles implanted is measured in terms of equivalent thickness (t_eq), defined as the thickness of a film made by an equivalent amount of nanoparticles deposited onto a rigid substrate. In this case, devices have been fabricated with t_eq = 81.5 ± 3 nm. It has been demonstrated that SCBI is an effective method for the microfabrication of stretchable conductive circuits on PDMS,⁵⁹ since it is applicable at room temperature (without any heating of the samples) and does not induce any charging or carbonization of the polymeric substrate.

As a following step, the nanoparticle paths are sealed to external wires by means of a layer of nickel graphite-filled silicone adhesive (Sol-08; Soliani EMC) cured at room temperature for 12 h. These soft contacts ensure both high conductivity and elasticity in the connecting region of the patterned surface to the external measuring circuitry (depicted in Fig. 1c). Finally, the whole surface is protected and electrically insulated with a 200-μm-thick film of soft elastomer (Ecoflex 00-30; Smooth-on, Inc.), as sketched in Figure 1d, deposited by spin coating at 500 rpm for 60 s and successively cured at room temperature for 12 h. The final device is shown in Figure 2a.

FIG. 2.

(a) Photograph of the final device. (b) Electrical schematization of the sensing system with its conditioning and read-out circuitry.

Read-out electronic system

The device is connected to an external custom-made printed circuit board (PCB), where the conditioning and the signal elaboration circuitry are embedded (see Figure 2b for the schematic electrical circuit) and the signal acquisition is performed by a PSoC3 (Programmable System on Chip CY8C3866AXI-040 from Cypress Semiconductor Corporation). Regarding the optical system, both the IR Led and the phototransistor are biased by means of variable resistors, to tune the emitting light intensity and the receiver collector voltage, respectively. Moreover, it is possible to vary the sensitivity of optical system. The phototransistor can be biased at different collector-to-emitter voltages V₀, and, at lower bias voltages, the sensitivity is higher (see Supplementary Data for additional details). In this way, the system would be able to adapt its response to different scenarios, where a lower or a higher sensitivity could be useful. To improve the system performance, a differential mode signal treatment is adopted. The photodiode is biased with a pulse wave, with a duty cycle of 2% at frequency of 20 Hz. For each cycle, the output signal is given by the difference between the collector voltage with and without the emitted diode light (i.e., before and at the end of the pulse wave biasing the diode, respectively).This strategy avoids loss of information and has two main advantages. First of all, the pulse length (1 ms) allows to filter the low-frequency noise of the diode current, such as the random telegraph signal noise usually due to trapping/detrapping phenomena, which affects the emitted light and, by consequence, the system output signal. The second advantage is the possibility to implement a differential read-out strategy to cancel the environmental light variations that may cause interferences to the output signal.

Regarding the resistive strain gauge, it is connected in a voltage divider with a 10 kΩ resistor and its signal is acquired by reading its voltage drop. In the case of the bending experiments with a cantilever configuration (shown in Supplementary Data), a capacitive linear accelerometer (LIS2L02AL; STMicroelectronics, Inc.) is connected to the same PCB. The output voltages, together with the accelerometer outputs, are collected and elaborated by a 32-bit PIC (PIC32MX460F512L; Microchip Technology, Inc.) microcontroller and transmitted to a PC by means of USB communication. Finally, the whole system is controlled by a custom-made graphical user interface.

Experimental setup

For characterizing the system response to external applied pressure, the device is fixed on a rigid surface, and it is indented in the middle of the nonimplanted (upper) surface, as depicted in Figure 3a. First, a Delrin^® probe with flat head (8 × 8 mm) is aligned to the center of the sensing area by means of three orthogonal manual micrometric translational stages with crossed roller bearing (M-105.10; PI); then, indentation of the loading probe in the normal direction is obtained by means of a servo-controlled micrometric translational stage (M-111.1; PI). At its opposite side, the probe is mechanically interfaced to a six-component load cell (ATI NANO 17 SI-25-0.25; Apex) capable of recording the indentation force applied to the soft sensing body (the acquisition frequency is 20 Hz). The probe is indented into the soft body at a constant velocity of 0.05 and 1 mm/s for quasistatic characterization and load/unload cycles, by gradually increasing the displacement and then the applied load.

FIG. 3.

Response to applied pressure. (a) Scheme of the experimental setup with the sensing system fixed on a rigid surface. The device is stimulated with a rigid flat square indenter, connected to a load cell, by means of a motorized vertical translation stage. (b) Output characteristics of the sensing system versus applied pressure. Different optical outputs are shown, with V₀ spanning from 0.5 to 2 V, demonstrating the tunable sensitivity of the system. In this case, the resistive output signal is null. (c) Output signals for load/unload cycles, with an applied pressure around 50 kPa and V₀ = 1 V.

To bend the device in a precise and repeatable way, it is installed on a thin metallic support by means of a Parafilm strip that ensures adhesion without affecting total flexibility. The metallic strip is clamped to a fixed mechanical support at one extremity, while the other end is clamped to a servo-controlled precision translation stage (M-126.CG1; PI) whose movement causes the bending of the metallic strip, and of the device fixed over it (see Fig. 4).

FIG. 4.

Response to negative/positive bending. (a) The device is fixed on a thin metallic strip of length L = 70 mm and bent by means of a horizontal translation stage connected to one extremity of the strip, with the other fixed. (b) Output characteristics of the optical (solid blue line) and resistive (red solid line) signals versus vertical deflection h. Schematics (c) and output signals (d) for load/unload cycles for negative bending, with vertical deflection around 5 mm. In this case, the optical signal is positive, while the resistive output is negative. Schematics (e) and output signals (f) for load/unload cycles for positive bending, with vertical deflection around −5 mm. In this case, both optical and resistive signals are positive.

The two previous experimental setups have been combined to perform tests by stimulating the device with outer pressure and bending at the same time. In particular, the setup used for pressure characterization, composed by the load cell and the flat indenter mounted on the precision vertical stage, is positioned above the bending system described above, as sketched in Figure 5a.

FIG. 5.

Sequential stimuli detection. (a) Output signals under sequential mechanical stimulation. In phase (I), the device is bent, without hitting the indenter placed above it. During this phase, both optical (blue solid line) and resistive (red solid line) signals vary. In phase (II), the indenter is moved cyclically in the vertical direction, applying a load/unload stimulation on the device. In this case, only the optical signal has a variation, in phase with the pressure measured by the load cell (black dashed line). Finally, in phase (III) the system is unbent and the initial flat configuration is retrieved. (b) Combined output signal (green solid line), proportional to the measured pressure (black dashed line).

Results and Discussion

The application of pressure causes a positive variation of the phototransistor output signal, while a null resistance variation of the gold nanoparticle paths is observed. The results are shown in Figure 3b, with a pressure up to 250 kPa. The optical output signal is given by the variation of the collector to emitter voltage ( \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${V_{CE}}$$ \end{document} ) of the phototransistor. In particular, we consider the relative voltage variation \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${ \rm{ \Delta }}V / {V_0}$$ \end{document} with \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${V_0} = {V_{CE}}$$ \end{document} . In the same plot, different signal variations of the optical system are depicted, with V₀ tuned from 0.5 to 2 V. As shown, the signal amplitude is much larger for lower values of V₀, demonstrating the possibility to adjust the sensitivity in a relevant way. While the configuration with \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${V_0} = 0.5 \,{ \rm{V}}$$ \end{document} is the most efficient, higher bias voltages could be used if lower outer pressures had to be filtered. For instance, an actuator could be triggered at a certain pressure level without introducing any other electronic/logical element, whose integration could be difficult in a soft robot. Moreover, in Figure 3c, the typical time responses of both signals following load/unload cycles are shown. For this example, the bias voltage is put to \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${V_0} = 1 \;{ \rm{V}}$$ \end{document} , and the indenter is cyclically loaded and unloaded, simulating a pressure square function and applying for each cycle a maximum pressure of 50 kPa.

In the case of bending, both optical and resistive signals vary. When the bending (i.e., positive) stretches the implanted surface, the resistance variation is positive; instead, in the case the same surface is compressed by an opposite bending (i.e., negative), the variation is negative (Fig. 4). Differently, the variation of the optical signal is always positive, since any mechanical deformation induces a decrease of the incident light on the receiver. Conventionally, positive or negative bending is caused by a positive or negative moment, respectively, while the vertical deflection has the opposite sign. For small deflections, the curved profile can be approximated with a circular arc. Given the strip length L = 70 mm and the vertical deflection h, it is possible to find the strain \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$\varepsilon$$ \end{document} as \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \varepsilon = - \frac { { \bar y } } { r } \simeq - { \frac { 4 \bar y } { { L^2 } } } h = - { \frac { 2t } { { L^2 } } } h , \tag { 1 } \end{align*} \end{document}

where \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$\bar y = t / 2$$ \end{document} is the vertical distance of the neutral axis to the implanted surface. Since the device thickness t is much higher with respect to those of the strip, the neutral axis is positioned near the middle of the device. Noticeably, while the relationship between the strain and the lateral compression \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${ \rm{ \Delta }}x$$ \end{document} is highly nonlinear (see Supplementary Data for detailed calculations), the strain versus vertical deflection characteristic becomes linear. Finally, considering the relative resistance variation \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${ \rm{ \Delta }}R / {R_0} = \alpha \varepsilon ,$$ \end{document} \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} { \frac { { \rm { \Delta } } R } { { R_0 } } } = - 2 \alpha \frac { t } { { { L^2 } } } h , \tag { 2 } \end{align*} \end{document}

the quasistatic characteristics of output signals versus vertical deflection h are shown in Figure 4b. The optical signal variation (blue solid curve) is positive for both positive and negative vertical deflections, while the resistance variation (red solid curve) follows Equation (2) (black dashed line), with \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${ \alpha _p} \simeq 24$$ \end{document} and \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${ \alpha _n} \simeq 12.5$$ \end{document} for positive and negative bending, respectively. These values represent also the gauge factors for positive and negative strain, respectively. This different value can be due to the different rearrangement of nanoparticle clusters during compression or extension. Figure 4d and f shows the time response of output signals due to bending/unbending cycles. In particular, Figure 4d shows the response to negative bending cycles. For each cycle, the system starts from the flat configuration (depicted in Fig. 4a), it reaches a maximum vertical deflection of 5 mm, and then it comes back to the initial flat condition. Instead, Figure 4f shows the response to positive bending cycles, where at each cycle the maximum deflection is of −5 mm. It can be noted that the optical signal has a positive variation in both cases, while the resistive variation is positive or negative according to the bending direction. Finally, for analyzing the hysteresis in bending/unbending cycles, the behavior of output signals versus vertical deflection is shown in Supplementary Data (see Supplementary Figs. S3 and S4 for negative and positive bending/unbending cycles, respectively). Also, a good repeatability and a fast response to the mechanical stimulation are observed.

The multimodal capabilities of the system are tested, as shown in Figure 5, to evaluate the possible application in soft robotics. First of all, the capability of distinguishing sequential stimulations is tested, and the result is shown in Figure 5b. In phase I, the sample, starting from a flat configuration, is bent negatively, without coming in contact with the indenter placed over it. This bending causes an increase of the optical output signal (blue solid line) and a negative resistance variation (red solid line). Then, in phase II, the indenter is cyclically moved vertically upward and downward, stimulating the bent device with load/unload cycles. The applied pressure is measured by the load cell (shown in black dashed line). As expected, during this phase, the optical signal presents an additional variation, in phase with the applied pressure, while the resistance change is null. Finally, in phase III, the cyclical vertical indentation is stopped, and the sample is released from bending, recovering initial values for both output signals. Observing this behavior, the two signals can be combined in such a way to enhance only the variation due to the external pressure, eliminating the contribution due to bending. In particular, by defining \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} S = { \frac { { \rm { \Delta } } V } { { V_0 } } } + k { \frac { { \rm { \Delta } } R } { { R_0 } } } , \tag { 3 } \end{align*} \end{document}

it is possible to find the optimal k value to have an almost null value for S in case of pure bending. The combined output signal, with \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$k = 2.3$$ \end{document} , is shown in Figure 5b (green solid line). The signal S has a positive variation only when the external force (black dashed line) is applied. Instead, as already shown in Equation (2), the bending is proportional only to the resistance variation. Hence, using Equation (2) and (3), the two different mechanical stimulations can be clearly distinguished and measured. Signal S is a linear combination of optical and resistive variations. Of course, this is an empirical choice, justified only by its effectiveness in giving a response proportional only to the applied pressure.

To further demonstrate the effectiveness of this strategy, also simultaneous mechanical stimulations are investigated. The results are shown in Figure 6. In this case, the indenter is put close to the sample surface. Then, when performing bending/unbending cycles, the sample hits the indenter (connected to the load cell) for a fraction of each cycle. In the noncontact phase, both output signals vary, as shown in Figure 6a. Instead, in the contact phase, the optical signal presents a very large variation (blue solid line), while the resistance change is almost null, being affected only by bending. The corresponding combined output signal (calculated with the same previous value for k) is shown in Figure 6b (green solid line), and it matches the pressure measured by the load cell (black dashed line). Then, also in this case, by exploiting Equation (2) and (3) it is possible to discriminate and measure bending and pressure.

FIG. 6.

Simultaneous stimuli detection. (a) Output signals under simultaneous mechanical stimulation. Following bending/unbending cycles, the device hits the indenter for a fraction of each cycle. In noncontact phase (I), the device approaches the indenter placed above it. Here, both optical (blue solid line) and resistive (red solid line) signals vary. In contact phase (II), the device hits the indenter, causing a pressure on its upper surface. In this case, only the optical signal has a variation, in phase with the pressure measured by the load cell (black dashed line). (b) Combined output signal (green solid line), proportional to the measured pressure (black dashed line).

Conclusion

In this work, we presented a multimodal mechanosensing system where the sensing elements are embedded in a soft body during its initial molding. Using the SCBI process, a strain gauge is incorporated without changing the mechanical properties of the bulk material (in this case PDMS). In particular, we demonstrate how, combining optical and the resistive outputs of two types of transducers in a proper (but very simple) way, the proposed system can detect and distinguish different stimuli, such as strain related to bending and external pressure. We believe that this technology has a high potential for building future multimodal soft robots with exteroception/proprioception. Indeed, to our knowledge, the only example of soft robot with these capabilities was presented by Larson et al.¹⁴ In that case, a unique electroluminescent crawling soft robot was implemented with pneumatic actuation and a hyperelastic skin. It was demonstrated how either pressure or the degree of the actuation could be retrieved with the embedded capacitive sensing. We believe that a new generation of soft robots capable of discriminating between the two sensing modalities could be developed if the principles and technologies that we demonstrate in this work are used and experimented further.

The aim of this work was to demonstrate the effectiveness of a novel strategy for multimodal sensing in soft robotics. However, several aspects remain to be properly studied and analyzed. In particular, future works should focus on the following aspects:

• How the responses of both electro and optical mechanisms would change in the case of integration in an entirely soft body, or if a pneumatic actuator is embedded in a soft robot or in a gripper.

• The dependence of the output signals on the characteristics of a contacting object. Indeed, in the case of indenters with different shapes, dimensions, and materials, the optical versus pressure characteristics will vary. Different shapes and dimensions will vary the contact area and, by consequence, the whole deformation of the soft waveguide. Also, a different material will modify the refraction of the IR wave inside the PDMS block. All these parameters will affect the wave intensity detected by the receiver. Then, a comprehensive work supported both by experiments and numerical calculations should quantify the influence of each of them, evaluating also the possible limitations.

• Finally, here we considered movements only in a 2D plane, while soft robots will move in the 3D space. Then, this sensing strategy should be generalized to reconstruct 3D deformations and movements.

Given the promising results at component level, the presented approach can be further investigated in the near future at robotic level to evaluate its potential during the movement and exploration of a soft robot, without needing unwieldy wiring and distributed conditioning circuits that nowadays are some of the main technological limits in this field.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

Rus

, Tolley

. Design, fabrication and control of soft robots. Nature, 2015; 521:467–475.

Laschi

. Soft robotics research, challenges, and innovation potential, through showcases (Soft Robotics, Transferring Theory to Application) Verl

, Albu-Schäffer

, Brock

, Raatz

. (Eds). Springer Berlin Heidelberg, 2015, pp. 255–264.

Paik

. Soft components for soft robots (Soft Robotics, Transferring Theory to Application) Verl

, Albu-Schäffer

, Brock

, Raatz

. (Eds). Springer Berlin Heidelberg, 2015, p. 272.

, Kim

D-H

. Flexible and stretchable electronics paving the way for soft robotics. Soft Robotics, 2014; 1:53–62.

Shepherd

, Ilievski

, Choi

, Morin

, Stokes

, Mazzeo

, et al. Multigait soft robot. Proc Natl Acad Sci U S A, 2011; 108:20400–20403.

Suzumori

, Endo

, Kanda

, Kato

, Suzuki

. A bending pneumatic rubber actuator realizing soft-bodied manta swimming robot. Proceedings 2007 IEEE International Conference on Robotics and Automation, April 10–14, 2007, p. 4975.

Trimmer

. Soft robots. Curr Biol, 2013; 23:R639–R641.

Robert

, Webster

, Jones

. Design and kinematic modeling of constant curvature continuum robots: a review. Int J Robot Res, 2010; 29:1661–1683.

Walker

, Dawson

, Flash

, Grasso

, Hanlon

, Hochner

, et al. Proc. SPIE 5804, Unmanned Ground Vehicle Technology VII, (May 27, 2005); p. 303.

10.

Kim

, Laschi

, Trimmer

. Soft robotics: a bioinspired evolution in robotics. Trends Biotechnol, 2013; 31:287–294.

11.

Iida

, Laschi

. Soft robotics: challenges and perspectives. Proc Comput Sci, 2011; 7:99–102.

12.

Beccai

, Lucarotti

, Totaro

, Taghavi

. Soft robotics mechanosensing. Soft Robotics: Trends, Applications and Challenges: Proceedings of the Soft Robotics Week, April 25–30, 2016, Livorno, Italy. Laschi

, Rossiter

, Iida

, Cianchetti

, Margheri

. (Eds). Cham: Springer International Publishing, 2017, p. 11.

13.

Case

, White

, Kramer

. Soft material characterization for robotic applications. Soft Robotics, 2015; 2:80–87.

14.

Larson

, Peele

, Li

, Robinson

, Totaro

, Beccai

, et al. Highly stretchable electroluminescent skin for optical signaling and tactile sensing. Science, 2016; 351:1071–1074.

15.

, Zhang

, Jia

, Mathewson

, Jang

K-I

, Kim

, et al. Soft microfluidic assemblies of sensors, circuits, and radios for the skin. Science, 2014; 344:70–74.

16.

Park

Y-L

, Chen

B-R

, Wood

. Design and fabrication of soft artificial skin using embedded microchannels and liquid conductors. IEEE Sensors J, 2012; 12:2711–2718.

17.

Laflamme

, Kollosche

, Connor

, Kofod

. Soft capacitive sensor for structural health monitoring of large-scale systems. Struct Control Health Monit, 2012; 19:70–81.

18.

Majidi

, Kramer

, Wood

. A non-differential elastomer curvature sensor for softer-than-skin electronics. Smart Mater Struct, 2011; 20:7.

19.

Mannsfeld

SCB

, Tee

BCK

, Stoltenberg

, Chen

CVHH

, Barman

, Muir

BVO

, et al. Highly sensitive flexible pressure sensors with microstructured rubber dielectric layers. Nat Mater, 2010; 9:859–864.

20.

Amjadi

, Pichitpajongkit

, Lee

, Ryu

, Park

. Highly stretchable and sensitive strain sensor based on silver nanowire–elastomer nanocomposite. ACS Nano, 2014; 8:5154–5163.

21.

Wang

, Jiu

, Nogi

, Sugahara

, Nagao

, Koga

, et al. A highly sensitive and flexible pressure sensor with electrodes and elastomeric interlayer containing silver nanowires. Nanoscale, 2015; 7:2926–2932.

22.

Jang

, Park

, Chen

, Das

, Kim

M-S

, Ahn

J-H

. Graphene-based flexible and stretchable electronics. Adv Mater, 2016; 28:4184–4202.

23.

Jeong

, Park

, Jin

, Hong

, Lee

S-S

, Ha

. Highly stretchable and sensitive strain sensors using fragmentized graphene foam. Adv Funct Mater, 2015; 25:4228–4236.

24.

Yan

, Wang

, Kang

, Cui

, Wang

, Foo

, et al. Highly stretchable piezoresistive graphene–nanocellulose nanopaper for strain sensors. Adv Mater, 2014; 26:2022–2027.

25.

Lipomi

, Vosgueritchian

, Tee

BCK

, Hellstrom

, Lee

, Fox

, et al. Skin-like pressure and strain sensors based on transparent elastic films of carbon nanotubes. Nat Nano, 2011; 6:788–792.

26.

Wichmann

MHG

, Buschhorn

, Böger

, Adelung

, Schulte

. Direction sensitive bending sensors based on multi-wall carbon nanotube/epoxy nanocomposites. Nanotechnology, 2008; 19:475503.

27.

Gong

, Schwalb

, Wang

, Chen

, Tang

, Si

, et al. A wearable and highly sensitive pressure sensor with ultrathin gold nanowires. Nat Commun, 2014; 5:3132 (2014).

28.

Sekitani

, Zschieschang

, Klauk

, Someya

. Flexible organic transistors and circuits with extreme bending stability. Nat Mater, 2010; 9:1015–1022.

29.

Someya

, Kato

, Sekitani

, Iba

, Noguchi

, Murase

, et al. Conformable, flexible, large-area networks of pressure and thermal sensors with organic transistor active matrixes. Proc Natl Acad Sci U S A, 2005; 102:12321–12325.

30.

Schwartz

, Tee

BCK

, Mei

, Appleton

, Kim

, Wang

, et al. Flexible polymer transistors with high pressure sensitivity for application in electronic skin and health monitoring. Nat Commun, 2013; 4:1859.

31.

Viry

, Levi

, Totaro

, Mondini

, Mattoli

, Mazzolai

, et al. Flexible three-axial force sensor for soft and highly sensitive artificial touch. Adv Mater, 2014; 26:2659–2664.

32.

Castano

, Flatau

. Smart fabric sensors and e-textile technologies: a review. Smart Mater Struct, 2014; 23:053001.

33.

Syduzzaman

, Patwary

, Farhana

, Ahmed

. Smart textiles and nano-technology: a general overview. J Textile Sci Eng, 2015; 5:181.

34.

Chou

H-H

, Nguyen

, Chortos

, To

JWF

, Lu

, Mei

, et al. A chameleon-inspired stretchable electronic skin with interactive colour changing controlled by tactile sensing. Nat Commun, 2015; 6:8011 (2015).

35.

Park

, Jang

. Fabrication of graphene/free-standing nanofibrillar PEDOT/P(VDF-HFP) hybrid device for wearable and sensitive electronic skin application. Carbon, 2015; 87:275–281.

36.

Park

, Lee

, Hong

, Lee

, Ha

, Jung

, et al. Tactile-direction-sensitive and stretchable electronic skins based on human-skin-inspired interlocked microstructures. ACS Nano, 2014; 8:12020–12029.

37.

Kim

D-H

, Lu

, Ma

, Kim

Y-S

, Kim

R-H

, Wang

, et al. Epidermal electronics. Science, 2011; 333:838–843.

38.

Shao

, Niu

, Hirtz

, Jiang

, Liu

, Wang

, et al. High-performance and tailorable pressure sensor based on ultrathin conductive polymer film. Small, 2014; 10:1466–1472.

39.

Webb

, Bonifas

, Behnaz

, Zhang

, Yu

, Cheng

, et al. Ultrathin conformal devices for precise and continuous thermal characterization of human skin. Nat Mater, 2013; 12:938–944.

40.

Kaltenbrunner

, Sekitani

, Reeder

, Yokota

, Kuribara

, Tokuhara

, et al. An ultra-lightweight design for imperceptible plastic electronics. Nature, 2013; 499:458–463.

41.

Yuen

, Cherian

, Case

, Seipel

, Kramer

. Conformable actuation and sensing with robotic fabric. Intelligent Robots and Systems (IROS 2014), 2014 IEEE/RSJ International Conference, September 14–18, 2014, p. 580.

42.

Leonov

. Thermoelectric energy harvesting of human body heat for wearable sensors. IEEE Sensors J, 2013; 13:2284–2291.

43.

Nesenbergs

, Selavo

. Smart textiles for wearable sensor networks: review and early lessons. Medical Measurements and Applications (MeMeA), 2015 IEEE International Symposium on IEEE, 2015, p. 402.

44.

Kong

J-H

, Jang

N-S

, Kim

S-H

, Kim

J-M

. Simple and rapid micropatterning of conductive carbon composites and its application to elastic strain sensors. Carbon, 2014; 77:199–207.

45.

Lee

, Kwon

, Seo

, Shin

, Koo

, Pang

, et al. Conductive fiber-based ultrasensitive textile pressure sensor for wearable electronics. Adv Mater, 2015; 27:2433–2439.

46.

Gerratt

, Michaud

, Lacour

. Elastomeric electronic skin for prosthetic tactile sensation. Adv Funct Mater, 2015; 25:2287–2295.

47.

Kim

, Park

, Jeong

, Kim

. Highly sensitive and multimodal all-carbon skin sensors capable of simultaneously detecting tactile and biological stimuli. Adv Mater, 2015; 27:4178–4185.

48.

Park

, Lee

, Hong

, Ha

, Jung

Y-D

, Lim

, et al. Giant tunneling piezoresistance of compositeelastomers with interlocked microdome arrays for ultrasensitive and multimodal electronic skins. ACS Nano, 2014; 8:4689–4697.

49.

Engel

, Nannan

, Tucker

, Chang

, Sung-Hoon

, Jones

. Flexible multimodal tactile sensing system for object identification, sensors, 2006. 5th IEEE Conference, October 22–25, 2006, p. 563.

50.

Lucarotti

, Totaro

, Sadeghi

, Mazzolai

, Beccai

. Revealing bending and force in a soft body through a plant root inspired approach. Sci Rep, 2015; 5:8788.

51.

Cianchetti

, Renda

, Licofonte

, Laschi

. Sensorization of continuum soft robots for reconstructing their spatial configuration. Biomedical Robotics and Biomechatronics (BioRob), 2012 4th IEEE RAS & EMBS International Conference, June 24–27, 2012, p. 634.

52.

Chossat

J-B

, Shin

H-S

, Park

Y-L

, Duchaine

. Soft tactile skin using an embedded ionic liquid and tomographic imaging. J Mechanisms Robotics, 2015; 7:021008.

53.

Park

, Kim

, Vosgueritchian

, Cheon

, Kim

, Koo

, et al. Stretchable energy-harvesting tactile electronic skin capable of differentiating multiple mechanical stimuli modes. Adv Mater, 2014; 26:7324–7332.

54.

, Sun

, Zhang

F-R

, Zhang

, Shi

L-A

, Zhao

H-Y

, et al. A stretchable electronic fabric artificial skin with pressure-, lateral strain-, and flexion-sensitive properties. Adv Mater, 2016; 28:722–728.

55.

Ramuz

, Tee

BCK

, Tok

JBH

, Bao

. Transparent, optical, pressure-sensitive artificial skin for large-area stretchable electronics. Adv Mater, 2012; 24:3223–3227.

56.

Levi

, Piovanelli

, Furlan

, Mazzolai

, Beccai

. Soft, transparent, electronic skin for distributed and multiple pressure sensing. Sensors, 2013; 13:6578–6604.

57.

Zhao

, O'Brien

, Li

, Shepherd

. Optoelectronically innervated soft prosthetic hand via stretchable optical waveguides. Sci Robotics, 2016; 1:eaai7529.

58.

Corbelli

, Ghisleri

, Marelli

, Milani

, Ravagnan

. Highly deformable nanostructured elastomeric electrodes with improving conductivity upon cyclical stretching. Adv Mater, 2011; 23:4504–4508.

59.

Borghi

, Melis

, Ghisleri

, Podestà

, Ravagnan

, Colombo

, et al. Stretchable nanocomposite electrodes with tunable mechanical properties by supersonic cluster beam implantation in elastomers. Appl Phys Lett, 2015; 106:121902.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.30 MB

0.00 MB