Smart Pneumatic Artificial Muscle Actuator withEmbedded Microfluidic Sensing

Yong-Lae Park1 and Robert J. Wood2,3

1Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213, USA.2Wyss Institute for Biologically Inspired Engineering, Harvard University, Boston, MA 02155, USA

3School of Engineering and Applied Science, Harvard University, Cambridge, MA 02138, USAE-mail: ylpark@andrew.cmu.edu, rjwood@eecs.harvard.edu

Abstract— We describe the design, fabrication, and charac-terization of a novel pneumatic artificial muscle actuator withembedded contraction sensing. The muscle is composed of threemain components: elastomer air chamber, embedded Kevlarthreads, and a helical microchannel filled with a liquid conductor.When the air chamber is inflated with compressed air, theconstrained length of the Kevlar threads causes the muscleto contract in the axial direction. During this contraction, themicrochannel can detect the shape change of the muscle bysensing, the expansion of the air chamber. This sensing capabilityincreases the controllability of pneumatic muscles. A novelmanufacturing method is proposed to embed Kevlar threads anda helical microchannel in an elastomer tube. Then, a liquid metalis injected into the microchannel to make a soft sensor thatcan detect the geometrical change of the muscle. The muscleprototype was characterized to demonstrate its actuation andsensing capability.

I. INTRODUCTION

Pneumatic artificial muscles (PAMs) have been widely usedin robotics and automation since they can easily generatelinear forces and displacements with a relatively compliant,compact, and lightweight form [1]. Examples of applicationsbased on their inherent compliance include human-friendlyrobots [2], robotic orthotic devices [3], [4], and soft wearablerobots [5], [6]. The operating principle of this type of actuatoris axial contraction of a flexible air chamber caused by radialexpansion of the chamber when filled with compressed air. Tocreate axial contraction motion and/or force, the air chamber isgeometrically constrained using different mechanisms. Someexamples include Mckibben muscles that enclose a stretchableair chamber with a braided mesh [7], [8], pleated muscles witha number of axial pleats directly fabricated on an air chambermembrane [9], [10], and fiber reinforced muscles that embedstraight fibers in a rubber air chamber [11], [12].

However, one typical limitation of these actuators is theirhighly non-linear behavior. This non-linearity makes it difficultto develop an accurate predictive model and consequently tocontrol for robotic applications. Furthermore, the addition ofexternal measurement devices for control, such as positionand force sensors, makes the entire mechanical system bulkyand complicated. Different approaches have been proposedto integrated sensing materials, such as conductive rubbertubes [13], conductive paste films [14], [15], and dielectric

Helical eGaIn microchannel

Embedded Kevlar threads

Inner layer

Outer layer Air chamber

eGaIn filled spiral microchannel Kevlar threads

Air inlet

(a)

(b)

40 mm

Fig. 1. (a) Multi-layered elastomer tube design with embedded Kevlar threadsand a helical microchannel. (b) Complete prototype.

polymers [16], for pneumatic and other soft actuators. How-ever, conductive rubber materials do not provide accurate andreliable measurements, and dielectric polymer materials cannotbe employed for large strains.

In this paper, we propose a smart artificial pneumatic muscleable to detect its contraction length by employing hyperelasticstrain sensing with embedded microchannels filled with a liq-uid conductor (e.g. eutectic Gallium Indium, eGaIn) [17], [18].The paper describes the design and fabrication of the prototypefollowed by characterization of the force, displacement, andsensor output. The prototype demonstrated highly linear andrepeatable sensor responses for measuring muscle contraction.

II. DESIGN

The base structure of the muscle is a double-layered elas-tomer tube (Fig. 1). The inner layer contains parallel Kevlarthreads. Compressed air injected into the air chamber expandsthe muscle in a radial direction and the constrained lengthof the Kevlar threads creates axial contraction of the muscle.

Kevlar thread

eGaIn microchannel

Outer layer Inner layer

1.7

elongated microchannel

Internal air pressure

(a) (b) Relaxed Pressurized eGaIn microchannel

Kevlar tread

Cross-­‐sec2onal views

Fig. 2. Muscle behavior and sensing principle. (a) Relaxed muscle. (b) Pres-surized muscle for axial contraction. The helical microchannel is elongateddue to the muscle’s radial expansion.

The outer layer contains an embedded helical microchannelfilled with eGaIn. The radial expansion of the muscle duringaxial contraction causes the elongation of the microchannel,resulting in increased electrical resistance of the microchannel(Fig. 2).

The resistance change of the microchannel is used to detectthe contraction length of the muscle. Since the microchan-nel is embedded outside the Kevlar thread layer, its shapechange is only dependent on the geometry change of theentire tube. Also, the helical microchannel design makes thesensor sensitive only to radial expansion of the muscle sinceaxial contraction alone does not change the length of themicrochannel.

III. FABRICATION

The muscle prototype was fabricated using a similar methodto the extended shape deposition manufacturing (SDM) pro-cess that was introduced for building sensor embedded ex-oskeletal robotic structures [19], [20]. In our process, 3Dprinted molds were used for casting an elastomer structureinstead of individually machined molds.

The base material of the muscle is silicone rubber1 (modu-lus: 150 kPa, shore A hardness: 10A), which was chosen forits stretchability (elongation at break: 1000%) and relativelyhigh tensile strength (3275 kPa).

1Dragonskin 10, Smooth-On, Inc.

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

Inner mold

Outer mold 1

Outer mold 2

Kevlar threads

Helical microchannel

Low-­‐fric;on fiber

Fig. 3. Fabrication Process. (a) Prepare molds (blue: cylindrical outer mold,brown: solid inner mold). (b) Place straight Kevlar threads. (c) Pour uncuredliquid elastomer. (d) Remove the outer mold when the elastomer cures. (e)Wrap the cured elastomer tube with a low-friction fiber in a helical shape.(f) Prepare a larger outer mold. (g) Pour another layer of uncured liquidelastomer. (h) Remover the outer and inner molds when the elastomer cures.(i) Remove the low-friction fiber by pulling for a helical microchannel.

The fabrication process can be divided into two major steps:inner layer casting with Kevlar threads and outer layer castingwith a helical microchannel.

In the first step, shown in Fig. 3(a)-(d), the outer and innermolds are prepared using a 3D printer2, and Kevlar threads3

(diameter: 350 µm) are embedded with an equally spacedstraight-line pattern. When the elastomer cures, the outer moldis removed, but the inter mold is kept in the elastomer tube.

In the second step, shown in Fig. 3(e)-(i), the cured elas-tomer tube with the inner mold is wrapped with a high-strength, low-friction fiber4 (diameter: 120 µm) in a helicalshape. Then, another elastomer layer is cast using a new outermold that is slightly larger than the first outer mold. Sincethe low-friction fiber is used to form an embedded helicalmicrochannel when it is pulled out of the cured elastomer,its diameter determines the size of the microchannel. Afterremoving both the outer and inner molds, eGaIn is injectedinto the helical microchannel.

2Objet500 Connex, Stratasys Ltd.38800K41, McMaster-Carr.4Spectra Braid Fishing Line, PEX.

10 mm

eGaIn microchannel

Kevlar thread

Relaxed

Pressurized

Contrac3on

Compressed air

Fig. 4. Muscle prototype relaxed (top) and pressurized at 70 kPa (bottom).The elongation of the microchannel is observed with muscle actuation.

IV. RESULTS

Fig. 1(b) shows the complete prototype, and Fig. 4 demon-strates actuation of the muscle prototype with compressed air.The relaxed length of the muscle is 40 mm. The outer diameterof the muscle is 9 mm, and the wall thickness of each layer is1 mm, which makes the inner diameter 5 mm. The diameterand length of the helical microchannel are approximately120 µm and 162 mm, respectively. A total of 20 Kevlarthreads were embedded at 18 intervals. The weight of thecomplete prototype including pneumatic fittings, metal crimps,and signal wires is 7.5 g.

The prototype was characterized for its actuation and sens-ing capability by applying different air pressures, as shownin Fig. 5. While applying varied air pressures, the contractionforce and length change of the muscle were measured using acommercial materials tester5. The air pressure was graduallyincreased up to 105 kPa. The prototype was able to createlinear contraction force and strain up to approximately 65 Nand 22.5%, respectively. The actuation behavior was highlynon-linear, as expected, but repeatable. During this experiment,the resistance change of the eGaIn sensor was also measured.The original resistance at rest was approximately 0.5 Ω,and with muscle contraction, the resistance increased up to5 Ω. The result showed high linearity and low hysteresis insensing muscle contraction with a sensitivity of approximately0.5 Ω/mm.

V. DISCUSSION

The main contribution of this work is the design of a sensor-embedded pneumatic artificial muscle and the developmentof a novel fabrication method. The current design providesposition sensing of muscle contraction requiring no externalsensors. When integrated into a robotic system for actuation,the proposed muscles will not only facilitate the control butalso make the mechanical system compact and simple.

Although the embedded position sensing capability of thecurrent design is highly useful for control, contraction forceis also critical information. Force sensing is currently being

5Instron 5544A, Instron.

0 1 2 3 4 5 6 7 8 9 10

70

60

50

40

30

20

10

0

Contrac3on (mm)

Force (N)

(a) 105 kPa 90 kPa 75 kPa 60 kPa 45 kPa

0 1 2 3 4 5 6 7 8 9 10 Contrac3on (mm)

6

5

4

3

2

1

0

Resistance change (Ω

)

(b)

Linear fit

Fig. 5. Prototype characterization results: (a) Actuation characterization.Shaded band represents the standard deviation of five trials for each pressurevalue. (b) Sensor characterization.

investigated by adding another layer containing a pressure-sensing eGaIn microchannel [18], [21]. The combination ofthe two sensing layers could be used to decouple positionand force information. Being equipped with both force andposition sensors in a compact form, this smart muscle willnot only expand the application areas of pneumatic artificialmuscles but also benefit an emerging field of soft robotics.

The proposed manufacturing method enables fabricationof a helical microchannel in a three-dimensional structure.However, the channel size is limited by the diameter of thelow-friction fiber. Improvements on the fabrication process byemploying different manufacturing approaches, such as liquidmetal patterning [22] and 3D printing [23], will facilitatefurther miniaturization of 3D microchannels.

Finally, in addition to the actuation and sensing capability,when miniaturized, the device may be distributed in a modularactuation system. Since the individual sensor-actuator unitcan be modularized, the collection of these modules willmake the system easily reconfigurable for different purposesand applications [24]. Furthermore, this modular design willallow the system to be easily resized without interrupting itsfunctionality and morphology.

VI. CONCLUSION

A smart pneumatic artificial muscle actuator was developed.The proposed muscle design contains an embedded sensingelement that can detect the length change of the muscle duringits contraction. For actuation, parallel Kevlar threads wereembedded in an elastomer tube to create axial contractionmotion and force with increased internal air pressure in thetube. For sensing, a helical microchannel filled with liquidmetal was embedded outside the Kevlar threads for detectingthe geometrical change of the muscle during actuation. Anovel manufacturing method was proposed for fabrication ofa helical microchannel embedded in a 3D elastomer structure.The characterization result showed linear and reliable sensorsignals.

ACKNOWLEDGMENT

This work was partially supported by the National ScienceFoundation (award number CNS-0932015) and the Wyss In-stitute for Biologically Inspired Engineering. Any opinions,findings, and conclusions or recommendations expressed inthis material are those of the authors and do not necessarilyreflect the views of the National Science Foundation.

REFERENCES

[1] F. Daerden and D. Lefeber, “Pneumatic artificial muscles: actuators forrobotics and automation,” Europ. J. Mech. Environ. Eng., vol. 47, pp.10–21, 2002.

[2] D. Shin, I. Sardellitti, Y.-L. Park, O. Khatib, and M. Cutkosky, “De-sign and control of a bio-inspired human-friendly robot,” InternationalJournal of Robotics Research, vol. 29, no. 5, pp. 571–584, 2010.

[3] D. P. Ferris, J. M. Czerniecki, and B. Hannaford, “An ankle-footorthosis powered by artificial peumatic muscles,” Journal of AppliedBiomechanics, vol. 21, pp. 189–197, 2005.

[4] B. G. Nascimento, C. B. S. Vimieiro, D. A. P. Nagem, and M. Pinotti,“Hip orthosis powered by pneumatic artificial muscle: Voluntary acti-vation in absence of myoelectrical signal,” Artificial Organs, vol. 32,no. 3, pp. 317–322, 2008.

[5] Y.-L. Park, B. Chen, D. Young, L. Stirling, R. J. Wood, E. Goldfield,and R. Nagpal, “Bio-inspired active soft orthotic device for anklefoot pathologies,” in Proc. IEEE/RSJ Int. Conf. Intell. Rob. Syst., SanFrancisco, CA, September 2011.

[6] Y.-L. Park, B. Chen, C. Majidi, R. Nagpal, and E. Goldfield, “Activemodular elastomer sleeve for soft wearable assistance robots,” in Proc.IEEE/RSJ Int. Conf. Intell. Rob. Syst., 2012, pp. 1595–1602.

[7] C.-P. Chou and B. Hannaford, “Measurement and modeling of McK-ibben pneumatic artificial muscles,” IEEE Trans. Rob. Autom., vol. 12,no. 1, pp. 90–102, 1996.

[8] B. Tondu and P. Lopez, “Modeling and control of McKibben artificialmuscle robot actuators,” IEEE Control Syst., vol. 20, no. 2, pp. 15–38,2000.

[9] B. Verrelst, R. V. Ham, B. Vanderborght, F. Daerder, M. V. Damme, andD. Lefeber, “Second generation pleated pneumatic artificial muscle andits robotic applications,” Adv. Rob., vol. 20, no. 7, pp. 783–805, 2006.

[10] F. Daerden and D. Lefeber, “The concept and design of pleated pneu-matic artificial muscles,” Int. J. Fluid Power, vol. 2, no. 3, pp. 41–50,2001.

[11] T. Nakamura and H. Shinohara, “Position and force control basedon mathematical models of pneumatic artificial muscles reinforced bystraight glass fibers,” in Proc. IEEE Int. Conf. Rob. Autom., 2007, pp.4361–4366.

[12] H. Tomori and T. Nakamura, “Theoretical comparison of McKibben-type artificial muscles and novel straight-fiber-type artificial muscles,”Int. J. Autom. Technol., vol. 5, no. 4, pp. 544–550, 2011.

[13] S. Kuriyama, M. Ding, Y. Kurita, J. Ueda, and T. Ogasawara, “Flexiblesensor for McKibben pneumatic artificial muscle actuator,” Int. J. Autom.Technol., vol. 3, no. 6, pp. 731–740, 2009.

[14] K. Kure, T. Kanda, K. Suzumori, and S. Wakimoto, “Flexible displace-ment sensor using injected conductuve paste,” Sens. Actuators, A, vol.143, no. 2, p. 272, 278 2008.

[15] Y. Yamamoto, S. Wakimoto, and K. Suzumori, “Evaluation of electroconductive film and strain gage as displacement sensor for pneumaticartificial muscle,” in Proc. IEEE Int. Conf. Rob. Biomimetics, 2011, pp.1206–1211.

[16] K. Jung, K. J. Kim, and H. R. Choi, “A self-sensing dielectric elastomeractuator,” Sens. Actuators, A, vol. 143, no. 2, pp. 343–351, 2008.

[17] Y.-L. Park, C. Majidi, R. Kramer, P. Berard, and R. J. Wood, “Hyperelas-tic pressure sensing with a liquid-embedded elastomer,” J. Micromech.Microeng., vol. 20, no. 12, 2010.

[18] Y.-L. Park, B. Chen, and R. J. Wood, “Design and fabrication of softartificial skin using embedded microchannels and liquid conductors,”IEEE Sens. J., vol. 12, no. 8, pp. 2711–2718, 2012.

[19] Y.-L. Park, K. Chau, R. J. Black, and M. R. Cutkosky, “Force sensingrobot fingers using embedded fiber Bragg grating sensors and shapedeposition manufacturing,” in Proc. IEEE Int. Conf. Rob. Autom., 2007,pp. 1510–1516.

[20] Y.-L. Park, S. C. Ryu, R. J. Black, K. Chau, B. Moslehi, and M. R.Cutkosky, “Exoskeletal force-sensing end-effectors with embedded op-tical fiber-bragg-grating sensors,” IEEE Trans. Rob., vol. 25, no. 6, pp.1319–1331, 2009.

[21] Y.-L. Park, D. Tepayotl-Ramirez, R. J. Wood, and C. Majidi, “Influenceof cross-sectional geometry on the sensitivity of liquid-phase electronicpressure sensors,” Appl. Phys. Lett., vol. 101, no. 19, 2012.

[22] R. K. Kramer, C. Majidi, and R. J. Wood, “Masked deposition ofgallium-indium alloys for liquid-embedded elastomer conductors,” Adv.Funct. Mater., 2013.

[23] C. Ladd, J.-H. So, J. Muth, and M. D. Dickey, “3D printing of freestanding liquid metal microstructures,” Adv. Mater., 2013.

[24] M. Yim, W.-M. Shen, B. Salemi, D. Rus, M. Moll, H. Lipson, E. Klavins,and G. S. Chirikjian, “Modular self-reconfigurable robot systems [grandchallenges of robotics],” IEEE Rob. Autom. Mag., vol. 14, no. 1, pp.43–52, 2007.