CNT-based sensor arrays for local strain measurements in soft pneumatic actuators

Regular Paper


Soft robotics is a recent trend in engineering that seeks to create machines that are soft, compliant, and capable of withstanding damage, wear and high stress. Soft pneumatic actuators (SPAs) are a key element of soft robots, and their elastomeric substrate enables generation of sophisticated motion with simple controls. Although several methods for fabrication, material selection, and structure design have been investigated for the construction of SPAs, limited attention has been paid to the integration of distributed sensors for performing localized measurement. Carbon nanotubes (CNTs) are molecular-scale tubes of carbon atoms with remarkable mechanical and electronic properties, showing potential application in sensing devices. In this paper, we present the design, fabrication, and testing of a novel type of CNT-based sensor array combined with silver nanowires for measuring localized strain along the bottom layer of a SPA. Simulation and experimentation have been performed in order to analyze the soft actuator deformation during bending. The results demonstrate the promise of the proposed SPA with integrated strain sensing, which lays groundwork for a myriad of applications in grasping, manipulation, and bio-inspired locomotion.


Soft robotics Strain sensor array CNT Integrated sensing 

1 Introduction

Traditionally, robots are composed of rigid bodies made from stiff materials such as metals and ceramics. These robots are widely applied in industry and can be preprogrammed to execute specific tasks with efficiency, but with constrained adaptability. The soft malleable state of elastomeric materials can enable a robot to readily adapt to different scenarios, facilitating tasks such as grasping (Brown et al. 2010) and improving mobility in unstructured and cluttered environments (McMahan et al. 2006). A soft robot is primarily made of soft and extensible materials such as silicone rubbers (Shepherd et al. 2011), synthetic fibers (Trivedi et al. 2008), or gels (Otake et al. 2002), enabling large deformation and absorption of energy generated from impacts. The highly deformable components allow the system to experience theoretically infinite degrees of freedom. In addition, the compliant structure mitigates the impact of environmental uncertainty and produces complex motion with simple control inputs. In contrast to their rigid counterparts, soft robots can often be fabricated with simple designs and inexpensive and lightweight materials.

The soft robotics literature is growing every day. These soft machines could contain single or multiple soft elements, such as bending actuators (Wakimoto et al. 2009), stretchable circuits (Rogers et al. 2010), and deformable sensors (Park et al. 2010). Soft robotic mechanisms can potentially be applied in many different areas, such as manufacturing (Amend et al. 2012), medicine (Cianchetti et al. 2014), assistive technologies (Polygerinos et al. 2015), and bio-inspired robotics (Laschi et al. 2012). However, an existing challenge is the integration of sensing devices in the structure of a soft mechanism, for feedback control of the system. Several control schemes for soft robotic systems have been explored by researchers (Duriez 2013; Marchese et al. 2014; Vikas et al. 2015) however, feedback control of soft robots using compact, integrated sensors is generally limited. A tactile sensor sleeve based on fiber-optic light modulation (Sareh et al. 2014) can be incorporated in the body of a soft manipulator for pressure sensing; however, large instrumentation devices are required for measuring the light intensity variation. Commercially available sensors can be embedded in a soft pneumatic actuator skin (Suh et al. 2014), but their size and shape cannot be customized for various applications. Although previous works have investigated the use of eGaIn-filled microchannels (Park 2013; Bilodeau et al. 2015) on soft actuators for measuring deformation, the type of sensor configuration investigated does not allow gathering of localized data about the deformed structure. In a more recent study (Morrow et al. 2016), multiple soft strain sensors made of eGaIn were integrated along a SPA structure, but this injecting process can be challenging and lead to entrapped air inside the sensing layer. By using sensors built in an integrated circuit (IC) the sensing accuracy can be increased (Ozel et al. 2016), but its rigid parts can compromise the flexibility of the soft-bodied device.

In this paper, we present the design, fabrication, and testing of a novel strain sensor array for measuring the deformation at different locations of a SPA. The actuator is fabricated with a soft lithography process, while the CNT-based sensor array and their AgNWs electrode traces are fabricated through a screen-printing process over a polymeric substrate. The screen-printing process has the advantage of being simple and scalable. The sensor design can be customized, characterized and used for detecting the soft actuator deformation. As a demonstration, a sensor array with three strain gauges was fabricated, and bonded to the bottom layer of a SPA. Different actuation tests were performed in order to analyze the sensor array response. The obtained sensor measurements showed good correlation with the actual SPA deformation.

The remainder of this paper is organized as follows. In Sect. 2, we describe the fabrication processes for both the soft actuator and flexible sensor devices. In Sect. 3, the characterization of the sensor array is presented as long with the finite element analysis of the soft actuation mechanism. Section 4 provides discussion based on the results obtained from this work, and concluding remarks are presented in Sect. 5.

2 Methodology

2.1 Fabrication of the soft actuator

The structure of a soft robot and its soft mechanisms can use several rubber-like materials. Similarly, there are different fabrication processes available for creating compliant robotic components such as soft lithography (Ilievski et al. 2011), 3D-printing (Peele et al. 2015), and photo-patterning (Breger et al. 2015). In particular, we have studied the fabrication of soft bending actuators using a soft lithography process.

A soft lithography process can be divided into three main steps: material preparation, vacuum degassing, and curing. The specific shape of the actuator is achieved through the design of a mold. Our mold was designed on a CAD software (SolidWorks, Dassault Systemes) and had a square cross-section of 25 × 20 mm and length of 100 mm. As shown in Fig. 1a, the SPA design included topographical features, a similar concept as in (Mosadegh et al. 2014), and with 70° angle to allow higher bending. The mold was made of polylactic acid (PLA) thermoplastic and fabricated with a 3D printer (MakerBot Replicator, MakerBot Industries). In order to facilitate the removal of the soft actuator after the curing process, we designed a 2-part mold with molded-in assembly features (Fig. 1b). We first mixed a two-part liquid silicone rubber (Dragon Skin 30, Smooth-On) and filled the mold with the uncured material. The silicone was then degassed in a vacuum chamber and allowed to cure for 16 h at the room temperature. Once the curing process was completed, the soft actuator was removed from the mold and its open ends were sealed with uncured silicone. In order to supply compressed air to the soft actuator chambers (see Fig. 1c), a silicone-based tubing with similar durometer (3.175 mm diameter) was inserted through the bottom layer of the actuator structure.
Fig. 1

Design approach for the soft actuator and mold parts. a Cross-section of the SPA with the corresponding dimensions; b core and cavity sides of the 3D printed mold; c inflation state of the SPA

2.2 Fabrication of the flexible sensor

In this work, we investigate new methods for achieving flexible sensors that can detect strains along a soft bending actuator. In Fig. 2, we demonstrate how a CNT-based sensor array can be fabricated through simple steps. This design allows the creation of customizable flexible sensors without geometric constraints.
Fig. 2

Fabrication steps of the flexible CNT-based sensor array: a the sensor mask is attached to the polymeric substrate; b the CNT conductive ink is applied over the mask surface to create the distributed strain sensors; c the trace mask is attached to the substrate; d AgNW solution is applied through the mask gaps using a pipette; e thin copper wires are attached to the endpoint of each trace with a silver paste; f encapsulation of the device with PDMS; g and bonding of SPA and sensor array with uncured silicone

An array of CNT-based strain gauges was fabricated using a screen-printing process. A polyimide film (Kapton, DuPont) was used for designing the pattern of the sensor array and circuit traces. In the first mask, we cut three equally spaced rectangles (10 × 2 mm) corresponding to the areas for the distributed sensor array along the device structure. A second mask was created to assist in the application of the trace material, with each sensor connected to two traces on its ends. The material used for the sensor substrate was polydimethylsiloxane (PDMS) (Sylgard 184, Dow Corning) with a 10:1 base and agent mix ratio. The substrate (≈2 mm thickness) was fabricated using two heat-resistant borosilicate glass sheets (150 × 150 mm) clamped together and heated over a hot plate for 10 min at 150 °C. A polyester film was adhered to the inner surface of each glass sheet to help in the removal of the PDMS substrate without causing any tear or wrinkles. The size and shape of the spacer material directly affect the substrate uniformity. In this procedure, we have used microscope slides with identical width and length to separate the glass sheets at opposite edges. After fabrication, the substrate was cut in a rectangular shape (120 × 25 mm) and treated with electric discharge to convert it to a less hydrophobic surface (wetting). The substrate and the first mask were attached together to allow the application of the sensor material (Fig. 2a). A single-walled CNT (SWCNT) conductive ink, 1 mg/mL CNTs (VC101, Chasm Technologies), was applied over the polyimide film (Fig. 2b), allowing the deposition of the SWCNTs on the PDMS substrate only through the small cuts. Small variations on ink dispersion can cause deviation in sensing characteristics. A spatula was used to level the CNT paste even with the polyimide film to obtain an ideal dispersion at each sensor location. The substrate was once again heated (60 °C) in order to ensure adhesion between both materials. Once dried, the second mask was attached to the substrate (Fig. 2c). AgNWs in water (AgNwL50H2O, ACS Material), with 50 nm of diameter and 200 µm of length, was applied through the open cuts using a pipette (see Fig. 2d). The traces were dried at 60 °C for 1 h. As shown in Fig. 2e, thin copper wires were attached to the endpoints of each trace by fixing them with a silver paste, 60% Ag (PELCO Colloidal Silver, Ted Pella). A second layer of uncured PDMS was applied to the top face of the sensor array to encapsulate the device (Fig. 2f).

As an example, the sensor array was fabricated with three strain gauges which are referred to as S1, S2 and S3 in this work, with S1 being closest to the air inlet, and S3 being close to the distal end of the actuator. The flexible sensor array was characterized (Sect. 3) and bonded to the bottom surface of the soft actuator using uncured PDMS (Fig. 2g). The fabricated devices are shown in Fig. 3.
Fig. 3

The fabricated devices. a SPA, b CNT-based sensor, and c both devices bonded together

2.3 Strain measurement

The CNT conductive ink has a mesh-like nanostructure which can change its electronic characteristics when subjected to extension or compression. When there is a voltage between both ends of an individual CNT-based sensor, the latter behaves as a variable resistor thathanges its resistance value during mechanical deformation. A voltage divider circuit can be used to measure the voltage across the flexible strain gauge sensor and thus its resistance value. By using Ohm’s Law, we have:
$$ \begin{array}{*{20}c} {V_{s} = I_{s} \times R_{s} } \\ \end{array} , $$
$$ \begin{array}{*{20}c} {I_{s} = \frac{{V_{i} }}{{R_{ref} + R_{s} }}} \\ \end{array} , $$
where \( V_{i} \), \( V_{s} \) are the input voltage and the voltage across the sensor, respectively, \( I_{s} \) is the current going through the sensor, and \( R_{s} \) and \( R_{ref} \). are the sensor resistance and the reference resistance, respectively. By combining (1) and (2), we can get,
$$ \begin{array}{*{20}c} {R_{s} = \frac{{R_{ref} }}{{\left( {\frac{{V_{i} }}{{V_{s} }} - 1} \right)}}} \\ \end{array} . $$
The resistance \( R_{ref} \) is selected based on the minimum and maximum resistance values obtained from each individual sensor (denoted as \( R_{s,min} \) and \( R_{s,max} \), respectively). This relation can be expressed as:
$$ \begin{array}{*{20}c} {R_{ref} \approx \sqrt {R_{s,min} \times R_{s,max} } } \\ \end{array} . $$
In order to associate the change in resistance with the amount of strain applied at a sensor position, multiple measurements of the resistance are recorded as the substrate is subjected to different values of strain. The axial strain formula is given by:
$$ \begin{array}{*{20}c} {\varepsilon = \frac{{\Delta L_{s} }}{{L_{so} }} = \frac{{L_{s} - L_{so} }}{{L_{so} }}} \\ \end{array} , $$
where Lso and \( L_{s} \) are the nominal (untensioned) length and the current length of the sensor array, respectively.

3 Results

3.1 Sensor characterization

The fabricated CNT-based sensor array was placed on a programmable stretching device with both ends clamped (80 mm active length). A loading cycle of ≈10 s (stretch and release) was performed for 60 min for the initial conditioning step of the device (Fig. 4). The sensor array was subjected to a stretch of 5% (\( L_{s} \) = 84 mm). This is an important process for stabilizing the microscale structural change of the nanomaterials. The conditioning phase was performed until the device achieved rersible and stable electrical signals for a desired range of strain. Although the sensor array had reached its stability margin after 60 min, the stretching cycle was run up to ≈180 min to ensure measurement repeatability. A data acquisition equipment was used in combination with an integrated dataflow software (LabVIEW, National Instruments) for collecting the change in resistance of each strain sensor.
Fig. 4

Continuous measurement of the resistance change in S1 during conditioning phase. The depicted lines show the difference in the maximum range of the sensor at initial step and after three time frames: 10, 30, and 60 min of loading cycle

After the conditioning step, each individual sensor was measured in a sequence. In Fig. 5, we show a short sample of the collected data from sensors \( S_{1} \) to \( S_{3} \) after 96 min of stretching. Each strain gauge experienced different ranges of resistance change since the concentration of the CNT material and the thickness along the substrate can directly affect their measurement range. The sensors fabricated in this work were tested only for stretches below 10%. A larger strain value could increase the number of cracks in the sensors, causing malfunction and unreliable measurements. Other design options that alleviate mechanical stress such as a horseshoe-shape would allow sensor functionality for larger stretches.
Fig. 5

Sensor array measurements after initial conditioning: a the amount of strain applied in the continuous stretch and release test, b and the relative change in resistance \( (\Delta R/R) \) for sensors S1, S2, and S3

Based on the data collected from the stretching process, we obtained the resistance-strain relation (gauge factor) for each individual CNT-based sensor (Fig. 6). As is observed, all three sensors achieve a saturation level at a certain strain value. Since the CNT bundles change their arrangement when the substrate is under strain, the conductive path within the nanomaterial is modified. The change in each sensor’s shape during stretching can also contribute to a saturated strain behavior. For small strain values, the gauge factor of each strain gauge is defined as:
$$ \begin{array}{*{20}c} {GF = \frac{\Delta R/R}{\varepsilon } = \frac{{\Delta R_{s} /R_{so} }}{{\Delta L_{s} /L_{so} }} = \frac{{\Delta R_{s} L_{so} }}{{R_{so} (L_{s} - L_{so} )}}} \\ \end{array} . $$
Fig. 6

Gauge factor of each strain gauge. a Strain and \( \Delta R/R \) relationship based on different sets of data (with mean lines) obtained through the cyclic stretching of the sensor array substrate, b and the standard deviation of \( \Delta R/R \) from the measured samples

3.2 Soft actuator simulation

Simulation of soft actuators and flexible sensors is an important process in the study of soft robotic systems, in order to analyze their mechanical behavior, frequency response and design performance. A Finite Element Method (FEM) model captures the nonlinearity of the system, but its complexity can lead to high computational cost. We have simulated several soft bending actuator prototypes, using a finite element analysis software (Abaqus/CAE, Dassault Systemes). The geometries of the soft actuator and the sensor substrate were designed as solid bodies and meshed using solid tetrahedral quadratic hybrid elements (C3D10H element type), with 48,253 nodes and 30,144 elements. The Dragon Skin 30 and PDMS materials were modeled as an incompressible Yeoh material (µ = 2.38 kPa) (Elsayed et al. 2014) and incompressible Neo-Hookean material (µ = 1.84 MPa) (Martinez et al. 2013), respectively. As shown in Fig. 7a, the simulated soft bending actuator achieved a quarter bending when a pressure of 40 kPa (≈5.8 psi) was applied to its inner chamber. The nodes 4455, 4365 and 4275 at the bottom surface were selected as the position for the strain gauges \( S_{1} \), \( S_{2} \) and \( S_{3} \), respectively. The uniaxial strains at the nodes that corresponded to the locations of the strain sensors were measured in the converged solution (Fig. 7b).
Fig. 7

Finite element simulation of the SPA. aContour plot of the SPA FEM model (DragonSkin/PDMS), and its sequential deformation as pressure increases. bGraph of the strain values at each node corresponding to a sensor location

3.3 Actuator-sensor testing

The performance of the integrated soft actuator-sensor was evaluated in experiments. In this procedure, the sensor S2 was not included since the process of embedding the sensor array substrate with the SPA caused a malfunction of this sensor. A pressure gauge sensor (ASDX Series, Honeywell) was used for detecting the inner pressure of the soft pneumatic actuator during the bending motion. The soft actuator inlet was connected to a 12VDC miniature air compressor (\( P_{max} \) = 28 psi) through a polyurethane tubing. During this test, we measured the resistance change of \( S_{1} \) and \( S_{3} \). When the sensor array was combined with the soft actuator, the minimum and maximum resistance values were changed due to the contribution of residual strain from the soft structure. By measuring \( R_{s1} \) and \( R_{s3} \) with a multimeter during the actuator inflation, we registered the new values as: \( R_{s1,min} \) ≈ 1.3 kΩ and \( R_{s1,max} \) ≈ 2.7 kΩ; \( R_{s3,min} \) ≈ 1.8 kΩ and \( R_{s3,max} \) ≈ 3.6 kΩ. From Eq. (4), we obtained the value for the reference resistors as \( R_{ref,s1} \) ≈ 1.8 kΩ, \( R_{ref,s3} \) ≈ 2.4 kΩ, and a series circuit was built using trim potentiometers.

The sensor outputs (pressure and strain gauges) were sent to a microcontroller (MEGA 2560, Arduino), which was connected to a computer using USB interface for communication with the data acquisition software. All measurements were obtained with a sampling time of ≈25 ms, and a solenoid valve connected to an air pump was controlled using pulse-modulation width (PWM) at 50 Hz. As shown in Fig. 8a, the fabricated soft actuator achieved 90° bending angle with an internal pressure of 60 kPa (≈8.7 psi), which is close to the FEM simulation results. The curvature of the soft actuator inner radius was captured with a digital camera, recording individual frames that correspond to a constant pressure value. The percentage of the PWM duty cycle was computed by a PID function within LabVIEW, with the process variable being the median of the pressure sensor output after thirty measurements. Various setpoints were tested with the control system, and the respective resistance values in each strain gauge were captured, as shown in Fig. 8b.
Fig. 8

The sequence of images of the SPA with embedded CNT-based sensors, showing the curvature and sensor values at different applied pressures. a Captured images during activation of the SPA, and b the strain gauge array measurement at various constant pressure setpoints

The SPA can be applied in several tasks such as reaching, grasping and touching. In order to have an accurate estimation about the actuator position or curvature, the sensor array must be able to capture the deformation during motion. In Fig. 9a, b, we show the measurements of the difference in the sensor output response (hysteresis) during inflation (loading) and deflation (unloading), with a range of pressure setpoints varying every 0.8 ms. The total operating time is 16 s, with the setpoint increasing from 0 to 60 kPa, and vice versa (Fig. 9c).
Fig. 9

Pressure and resistance relationship for the fabricated SPA with embedded sensor array. The variation in resistance (with mean lines) for both inflation (loading) and deflation (unloading) steps are shown for sensors S1 (a) and S3 (b), when the pressure setpoint was varying from 0 to 60 kPa. c A sequence of frames showing the actuation steps and the curvature of the SPA

With the gauge factor obtained during the sensor array characterization process, we estimated the amount of strain in the SPA at different curvatures. The actual curvature of the SPA was measured from the captured image frames during bending. An image processing software (Vision Assistant, National Instruments) was used to measure the SPA inner radius of curvature for each pressure setpoint. The arc length of a specific bending shape was measured and inserted in Eq. (5) (with \( L_{so} \) = 100 mm) to find its correspondent axial strain value (\( \varepsilon \)). In Fig. 10, we compare both sensor and actual measurements when the curvature increases. These results agree with the compression observed in the bottom layer of the SPA during positive pressure.
Fig. 10

The relationship between strain and curvature of the SPA based on the range of resistance values captured by the sensor array. The actual value represents the measured curvature from the SPA inner radius

4 Discussion

4.1 Soft actuator curvature

Many different silicone rubber materials can be used for fabricating soft actuators. An elastomeric material is characterized by its durometer level (Shore Hardness). Low durometer silicones can resist extensive elongations, allowing the soft actuator to generate full bending curvature at low pressure values. By implementing the soft bending actuator and the integrated sensors with the aforementioned materials, the soft bending actuator was able to generate a 90° bending angle with pressure values less than 60 kPa.

4.2 Sensor substrate thickness

The constraint layer of a soft bending actuator, which in this work was considered as the substrate of the sensor array, can directly affect the actuator performance (bending curvature) when its thickness value is changed. A thinner layer will enable the actuator to require less pressure in order to achieve full bending. On the other hand, decreasing the thickness of the substrate brings a challenge to the fabrication of the sensor device since the uncured silicone dispersion creates a nonuniform surface. This causes a drastic variation among the characterizations of each strain gauge in the sensor array. Moreover, it generates difficulty in predicting the gauge factor for each sensor, which removes the possibility of a single circuit design (same \( R_{ref} \)) for all sensors in a voltage divider configuration.

4.3 Sensor array

By integrating multiple strain gauges along the soft actuator structure, the sensor array can provide angular measurement, contact detection or proprioceptive sensing for a soft robotic system. The distributed measurements allow estimation of the actuator deformation and the locations and forces of interactions between the actuator and foreign objects. Since the sensors are fabricated through screen-printing process, customization can be applied according to the soft actuator geometry and strain directions of interest. In addition, the number of strain sensors and their position can be optimally chosen based on the regions which experience large deformations. The sensor array design presents some constraints regarding the number of sensors that can be fabricated in a single substrate. Since the trace width can affect the sensor reading, reducing its dimension to increase the amount of strain gauges can impact the reliability of the sensor measurement. Increasing the thickness of each trace may allow the design of thinner features, but this process was not investigated in this current work. The manual fabrication method applied in this work brings challenges for achieving a uniform substrate thickness and CNT ink dispersion. Automated mass production ultraviolet (UV) curing and screen-printing processes could reduce the uncertainties and issues identified in the investigated method.

5 Conclusions

We investigated the design and fabrication of CNT-based strain sensors that can be embedded in the structure of a soft robotic component. An array of strain gauges can be used as a distributed sensor network along the compliant mechanism. The sensor size and shape can be customized for many different applications. In this work, we tested the CNT-based strain sensor for detecting deformation at different locations at the bottom surface of a SPA. The results provide initial steps in the implementation of a sensor array for monitoring local deformation on a soft robotic mechanism.

There are several directions for extending the presented work. We are working towards new types of soft robotic actuators that can exploit sensing capabilities integrated in their mechanical structure. More investigation needs to be performed in order to achieve better uniformity on the sensor substrate and ink dispersion, which will facilitate the characterization process of multiple embedded sensors. While the present work focused on single-degree-of-freedom strain measurement, future work involves the study of sensor arrays for multi-axis measurements. Furthermore, the integrated strain gauges will be used in feedback control of the SPA to achieve accurate displacement and force control.



This work was supported by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) under the Science Without Borders program (BEX-13404-13-0) and National Science Foundation (DBI-0939454 and ECCS-1549888).

Supplementary material

41315_2017_18_MOESM1_ESM.mp4 (11.9 mb)
Supplementary material 1 (MP4 12184 kb)


  1. Amend, J.R., Brown, E., Rodenberg, N., Jaeger, H.M., Lipson, H.: A positive pressure universal gripper based on the jamming of granular material. IEEE Trans. Robot. 28(2), 341–350 (2012)CrossRefGoogle Scholar
  2. Bilodeau, R.A., White, E.L., Kramer, R.K.: Monolithic fabrication of sensors and actuators in a soft robotic gripper. In: Intelligent Robots and Systems (IROS), 2015 IEEE/RSJ International Conference on, Hamburg, Germany (2015)Google Scholar
  3. Breger, J.C., Yoon, C., Xiao, R., Kwag, H.R., Wang, M.O., Fisher, J.P., Nguyen, T.D., Gracias, D.H.: Self-folding thermo-magnetically responsive soft microgrippers. ACS Appl. Mater. Interfaces. 7(5), 3398–3405 (2015)CrossRefGoogle Scholar
  4. Brown, E., Rodenberg, N., Amend, J., Mozeika, A., Steltz, E., Zakin, M.R., Lipson, H., Jaeger, H.M.: Universal robotic gripper based on the jamming of granular material. Proc. Natl. Acad. Sci. 107(44), 18809–18814 (2010)CrossRefGoogle Scholar
  5. Cianchetti, M., Ranzani, T., Gerboni, G., Nanayakkara, T., Althoefer, K., Dasgupta, P., Menciassi, A.: Soft robotics technologies to address shortcomings in today’s minimally invasive surgery: the STIFF-FLOP approach. Soft Robot. 1(2), 122–131 (2014)CrossRefGoogle Scholar
  6. Duriez, C.: Control of elastic soft robots based on real-time finite element method. In: Robotics and Automation (ICRA), 2013 IEEE International Conference on, Karlsruhe, Germany (2013)Google Scholar
  7. Elsayed, Y., Vincensi, A., Lekakou, C., Geng, T., Saaj, C.M., Ranzani, T., Cianchetti, M., Menciassi, A.: Finite element analysis and design optimization of a pneumatically actuating silicone module for robotic surgery applications. Soft Robot. 1(4), 255–262 (2014)CrossRefGoogle Scholar
  8. Ilievski, F., Mazzeo, A.D., Shepherd, R., Chen, X., Whitesides, G.M.: Soft robotics for chemists. Angew. Chem. 123(8), 1930–1935 (2011)CrossRefGoogle Scholar
  9. Laschi, C., Cianchetti, M., Mazzolai, B., Margheri, L., Follador, M., Dario, P.: Soft robot arm inspired by the octopus. Adv. Robot. 26(7), 709–727 (2012)CrossRefGoogle Scholar
  10. Marchese, A.D., Komorowski, K., Onal, C.D., Rus, D.: Design and control of a soft and continuously deformable 2d robotic manipulation system. In: Robotics and Automation (ICRA), 2014 IEEE International Conference on, Hong Kong, China (2014)Google Scholar
  11. Martinez, R.V., Branch, J.L., Fish, C.R., Jin, L., Shepherd, R.F., Nunes, R., Suo, Z., Whitesides, G.M.: Robotic tentacles with three-dimensional mobility based on flexible elastomers. Adv. Mater. 25(2), 205–212 (2013)CrossRefGoogle Scholar
  12. McMahan, W., Chitrakaran, V., Csencsits, M., Dawson, D., Walker, I.D., Jones, B.A., Pritts, M., Dienno, D., Grissom, M., Rahn, C.D.: Field trials and testing of the OctArm continuum manipulator. In: Robotics and Automation, 2006. ICRA 2006. Proceedings 2006 IEEE International Conference on, Orlando, Florida (2006)Google Scholar
  13. Morrow, J., Shin, H.-S., Phillips-Grafflin, C., Jang, S.-H., Torrey, J., Larkins, R., Dang, S., Park, Y.-L., Berenson, D.: Improving soft pneumatic actuator fingers through integration of soft sensors, position and force control, and rigid fingernails. In: Robotics and Automation (ICRA), 2016 IEEE International Conference on, Stockholm, Sweden (2016)Google Scholar
  14. Mosadegh, B., Polygerinos, P., Keplinger, C., Wennstedt, S., Shepherd, R.F., Gupta, U., Shim, J., Bertoldi, K., Walsh, C.J., Whitesides, G.M.: Pneumatic networks for soft robotics that actuate rapidly. Adv. Func. Mater. 24(15), 2163–2170 (2014)CrossRefGoogle Scholar
  15. Otake, M., Kagami, Y., Inaba, M., Inoue, H.: Motion design of a starfish-shaped gel robot made of electro-active polymer gel. Robot. Auton. Syst. 40(2), 185–191 (2002)CrossRefGoogle Scholar
  16. Ozel, S., Skorina, E.H., Luo, M., Tao, W., Chen, F., Pan, Y., Onal, C.D.: A composite soft bending actuation module with integrated curvature sensing. In: Robotics and Automation (ICRA), 2016 IEEE International Conference on, Stockholm, Sweden (2016)Google Scholar
  17. Park, Y.L., Majidi, C., Kramer, R., Bérard, P., Wood, R.J.: Hyperelastic pressure sensing with a liquid-embedded elastomer. J. Micromech. Microeng. 20(12), 125029 (2010)CrossRefGoogle Scholar
  18. Park, Y.L., Wood, R.J.: Smart pneumatic artificial muscle actuator with embedded microfluidic sensing. In: SENSORS, 2013 IEEE, Baltimore, Maryland (2013)Google Scholar
  19. Peele, B.N., Wallin, T.J., Zhao, H., Shepherd, R.F.: 3D printing antagonistic systems of artificial muscle using projection stereolithography. Bioinspir. Biomim. 10(5), 055003 (2015)CrossRefGoogle Scholar
  20. Polygerinos, P., Galloway, K.C., Savage, E., Herman, M., O’Donnell, K., Walsh, C.J.: Soft robotic glove for hand rehabilitation and task specific training. In: Robotics and Automation (ICRA), 2015 IEEE International Conference on, Seattle, Washington (2015)Google Scholar
  21. Rogers, J.A., Someya, T., Huang, Y.: Materials and mechanics for stretchable electronics. Science 327(5973), 1603–1607 (2010)CrossRefGoogle Scholar
  22. Sareh, S., Jiang, A., Faragasso, A., Noh, Y., Nanayakkara, T., Dasgupta, P., Seneviratne, L.D., Wurdemann H.A., Althoefer, K.: Bio-inspired tactile sensor sleeve for surgical soft manipulators. In: Robotics and Automation (ICRA), 2014 IEEE International Conference on, Hong Kong, China (2014)Google Scholar
  23. Shepherd, R.F., Ilievski, F., Choi, W., Morin, S.A., Stokes, A.A., Mazzeo, A.D., Chen, X., Wang, M., Whitesides, G.M.: Multigait soft robot. Proc. Natl. Acad. Sci. 108(51), 20400–20403 (2011)CrossRefGoogle Scholar
  24. Suh, C., Margarit, J.C., Song, Y.S., Paik, J.: Soft pneumatic actuator skin with embedded sensors. In Intelligent Robots and Systems (IROS 2014), 2014 IEEE/RSJ International Conference on, Chicago, Illinois (2014)Google Scholar
  25. Trivedi, D., Dienno, D., Rahn, C.D.: Optimal, model-based design of soft robotic manipulators. J. Mech. Des. 130(9), 091402 (2008)CrossRefGoogle Scholar
  26. Vikas, V., Grover, P., Trimmer, B.: Model-free control framework for multi-limb soft robots. In: Intelligent Robots and Systems (IROS), 2015 IEEE/RSJ International Conference on, Hamburg, Germany (2015)Google Scholar
  27. Wakimoto, S., Ogura, K., Suzumori, K., Nishioka, Y.: Miniature soft hand with curling rubber pneumatic actuators. In: Robotics and Automation, 2009. ICRA ‘09. IEEE International Conference on, Kobe, Japan (2009)Google Scholar

Copyright information

© Springer Singapore 2017

Authors and Affiliations

  1. 1.Department of Electrical and Computer EngineeringMichigan State UniversityEast LansingUSA

Personalised recommendations