Bangyuan LiuCollege of Control Science and Engineering,

Zhejiang University,

38 Zheda Road,

Hangzhou 310027, China

e-mail: bennyliu@zju.edu.cn

Feiyu ChenCollege of Mechanical Engineering,

Zhejiang University,

38 Zheda Road,

Hangzhou 310027, China

e-mail: feiyuchen@zju.edu.cn

Sukai WangCollege of Biomedical Engineering and

Instrument Science,

Zhejiang University,

38 Zheda Road,

Hangzhou 310027, China

e-mail: wangsukai@zju.edu.cn

Zhiqiang FuCollege of Mechanical Engineering,

Zhejiang University,

38 Zheda Road,

Hangzhou 310027, China

e-mail: zhiqiangfu@zju.edu.cn

Tingyu ChengJonh A. Paulson School of Engineering and

Applied Sciences,

Harvard University,

Cambridge, MA 02138

e-mail: tingyucheng@g.harvard.edu

Tiefeng LiDepartment of Engineering Mechanics,

Key Laboratory of Soft Machines and

Smart Devices of Zhejiang Province,

Soft Matter Research Center (SMRC),

Zhejiang University,

38 Zheda Road,

Hangzhou 310027, China

e-mail: litiefeng@zju.edu.cn

Electromechanical Controland Stability Analysis of a SoftSwim-Bladder Robot Drivenby Dielectric ElastomerCompared to the conventional rigid robots, the soft robots driven by soft active materialspossess unique advantages with their high adaptability in field exploration and seamlessinteraction with human. As one type of soft robot, soft aquatic robots play important rolesin the application of ocean exploration and engineering. However, the soft robots stillface grand challenges, such as high mobility, environmental tolerance, and accurate con-trol. Here, we design a soft robot with a fully integrated onboard system including powerand wireless communication. Without any motor, dielectric elastomer (DE) membranewith a balloonlike shape in the soft robot can deform with large actuation, changing thetotal volume and buoyant force of the robot. With the help of pressure sensor, the robotcan move to and stabilize at a designated depth by a closed-loop control. The perform-ance of the robot has been investigated both experimentally and theoretically. Numericalresults from the analysis agree well with the results from the experiments. The mecha-nisms of actuation and control may guide the further design of soft robot and smartdevices. [DOI: 10.1115/1.4037147]

1 Introduction

Robots have played essential roles in modern society becauseof their high output force, controllability, and precision. However,most robots are developed based on hard materials, which havelimitations in energy utilization, adaptability, and human–robotinteraction. Soft robots attract growing attention [1–6] because oftheir unique advantages over conventional rigid robots, includinglarge actuation, high adaptability, and high compatibility withhuman. Aquatic creatures inspire the design of soft robots due totheir large fraction of soft body and high agility [7–12]. Theincreasing importance of ocean missions generates the demandsfor developing high-performance aquatic soft robot.

An increasing number of researchers focuses on developingunderwater robot systems for various missions, such as marine

environment monitoring, exploration of marine resource, andinvestigation of underwater creatures [13]. Conventional aquaticrobot driven by hard components such as electric motors and elec-tromagnetic actuators [14–18] has limitations in the field opera-tion due to their high noise and low adaptability. Soft activematerials, such as ionic conducting polymer film [19], shapememory alloys [20,21], and ionic polymer metal composites[22–24], can be used as artificial muscles to drive the aquatic softrobots. These robots have limitations because of the low energydensity [20,21], small deformation [20,21], and slow response[19,22–24]. Achieving fast response, high energy density, andlarge deformation, dielectric elastomer (DE) [25–27] stands outamong various stimuli-responsive materials. Recently, severalgroups have developed aquatic soft robots driven by dielectricelastomer. Zhu and coworkers [1] have designed a jellyfish robot,Li et al. [28] have designed a Manta-ray robot, and Andersonet al. [29] have designed an artificial muscle actuator for a roboticfish. Michel and coworkers [30] have designed fishlike propulsionof an airship.

Contributed by the Applied Mechanics Division of ASME for publication in theJOURNAL OF APPLIED MECHANICS. Manuscript received April 25, 2017; finalmanuscript received June 20, 2017; published online July 12, 2017. Assoc. Editor:Kyung-Suk Kim.

Journal of Applied Mechanics SEPTEMBER 2017, Vol. 84 / 091005-1Copyright VC 2017 by ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Those robots show good performance in fast response, largeactuation, and environmental adaptability. However, those robotsshow difficulty in controllability and robustness, and no controlalgorithm has been applied to any of them. In this paper, we focuson designing an aquatic soft robot driven by a bladder-shaped DEactuator. Onboard system for power, control, and communicationis developed. Control algorithm has been applied to tune the oper-ation depth and stabilize the movements of the robot.

This paper is arranged as follows: Section 2 has discussed themechanical design and actuation mechanism of the robot. Section3 presents the process and results of the underwater experiments.Section 4 derives an analytical model for the DE actuator. Theanalytical results are compared with the experimental measure-ments. Section 5 illustrates stability of the system with and with-out control. The performance of the robot has been evaluated.

2 Mechanical Design and Movement Mechanism of

the Robot

Inspired by the structure and floating mechanism of the swim-bladder, we design the artificial swim-bladder robot by usingDE membrane (3MTM VHBTM) as the artificial muscle. Thedetailed fabrication process is shown in Fig. 1. The artificial mus-cle laminate (Fig. 1(a)) consists of a thin DE membrane coatedwith carbon grease. We stack the 3MTM VHBTM 4910 and 4905membranes together to make DE membrane with various thick-nesses (0.5 mm, 1 mm, 1.5 mm, 2 mm, and 2.5 mm) in the experi-ments. Then, the DE membrane is biaxially prestretched (3� 3)and fixed on the open end of an acrylic chamber. The size of theacrylic chamber is 6 cm in diameter and 10 cm in height (Figs.1(b) and 1(c)). Then, we seal the chamber and add counterweightto balance the buoyant force (Fig. 1(d)). Air is pumped into thechamber deforming the DE membrane into a balloon shape(Fig. 1(e)).

Figure 2 shows the actuation mechanism of the swim-bladderrobot. While the carbon grease serves as electronic conductiveelectrode, the surrounding water is utilized as the ionic electrode.Despite the low electrical conductivity of the water (�50 mS/m),it is sufficient to serve as an effective electrode [28], which willbe evidenced later in the context. Without applying a voltage, theDE membrane is deformed by the pressure difference and

maintains an equilibrium state (the inflated state). Figures 2(a)and 2(c) show the inflated state in schematic and experimentalimage. When a voltage is applied (Figs. 2(b) and 2(d)), positiveand negative charges accumulate on both sides of the DE, induc-ing Maxwell stress [31] and deforming the DE membrane (theactuated state). The actuation of the DE membrane changes thevolume and the buoyant force of the DE balloon. As a result, theposition of the swim-bladder robot varies.

3 Experiments

The goal of this study is to achieve a controllable artificialswim-bladder robot with onboard power and control. The experi-mental setup includes a DE membrane, carbon grease, an acrylicbottle, an onboard power source which is named as “Epod,” and apressure sensor circuit (Fig. 3). The system in Epod is controlledby an eight-pin microcontroller unit (MCU) to tune the voltageamplitude. The high voltage amplitude is adjusted by pulse-widthmodulation duty cycle. Pressure sensor circuit is powered by alow voltage battery (3.7 V) with a pressure sensor connecting tothe MCU. Two circuits exchange data via two 2.4G ZigBee wire-less modules. Epod tunes the voltage amplitude according to thepressure signal sent by pressure sensor circuit. Besides, a com-puter can communicate with Epod via ZigBee in order to displayand monitor real time data. Figure 3(d) plots the experimentaldata of pressure increment as a function of voltage, which is col-lected by the pressure sensor.

The floating performance of the robot is affected by the initial vol-ume of the DE balloon, which can be tuned by pumping air into thechamber with an initial pressure. Figure 4(a) plots the experimentaldata of the buoyant force increment as a function of voltage withvarious initial pressures. The thickness of DE membrane is 2 mm.The depth is 25 cm. When the membrane starts actuation from ahemisphere shape without voltage (Fig. 4(e), and the initial innerpressure is about 108 kPa), the device achieves the largest variationof the buoyant force (voltage from 0 to 10 kV) than the other condi-tions with various initial shapes of the balloon (Figs. 4(b)–4(g)).

In the experiment, we performed a series of experiments toevaluate the performance of the swim-bladder robot with variousthicknesses of DE membrane (Fig. 5). Figure 5(a) plots the experi-mental data of volume increment as a function of voltage. Figure5(c) illustrates the experimental data of buoyant force increment

Fig. 1 Fabrication of controllable artificial swim bladder robot. (a) A DE membrane (stackedVHB membrane with the initial thickness of 2 mm) was biaxially prestretched (3 3 3) on theABS frame. (b) Carbon grease was coated with a circular shape on one side of the membrane.(c) The prestretched DE membrane was assembled on the top of the acrylic chamber with thecarbon grease coated side facing downward. (d) Seal the chamber and add counterweight tobalance the buoyant force. (e) The air is pumped into the chamber deforming the DE mem-brane into a balloon shape.

091005-2 / Vol. 84, SEPTEMBER 2017 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

as a function of voltage. The buoyant force was measured by aforce sensor connected to the swim-bladder robot (the force differ-ence before and after the actuation of the DE membrane). Underthe same voltage value, thinner membrane provides larger buoyantforce change before electrical breakdown (the black cross). How-ever, thicker membrane could endure higher voltage and couldfinally attain larger change of buoyant force. The maximum voltagevalue is limited under 10 kV due to the limitation of the instrument.

The change of buoyant force of the robot is affected by bothvoltage and depth. The change of buoyant force could be repre-sented by the total volume of the DE membrane and the chamber.Figure 5(e) plots the experimental data of total volume as a functionof both voltage and various depths. The results show that there isno interconnection between the voltage-induced volume expansionand depth-induced volume expansion. This behavior guaranteesthat the linear control is possible for the floating of the robot.

To further investigate the behaviors of the robot, we havedeveloped an analytical model (Sec. 4). To prove the validity ofthe model, we used MATLAB to solve the implicit function. Figures 5(b)and 5(d) show the volume–voltage relation and the buoyantforce–voltage relation with various thicknesses of DE membrane(corresponding to the experimental results in Figs. 5(a) and 5(c)).Figure 5(f) shows the analytical results of the pressure–voltagerelation with various depths (corresponding to the experimentalresults in Fig. 5(e)). The analytical results qualitatively agree withthe experimental measurements.

We use an independent driving power source to drive the robotin order to replace the bulky high voltage power source. Figure 6plots the experimental data of the voltage-induced volumeexpansion as a function of the control signal. There is a positivecorrelation between control signal and actual voltage. To keep theswim-bladder robot at a specific depth z, an on–off control strat-egy is used: the output is set either on or off according to the airpressure. When the output is turned on, the power source applies acontrol signal (with the value of 15,000). Before the control pro-cess, the system is calibrated with the following five steps:

(1) Place the robot at a certain place in the water tank with adepth of z0.

(2) Measure the air pressure Poff0 without applying voltage.(3) Measure the air pressure Pon0 after turning on the voltage.(4) Calculate the variation of the air pressure DP when the

robot is placed 1 cm deeper in the water.(5) Calculate two pressure values Pon and Poff (the desired

pressure when the robot stays at a depth of z).

Pon and Poff can be calculated by

Poff ¼ Poff0 þ ðz� z0Þ � DP (1)

Pon ¼ Pon0 þ ðz� z0Þ � DP (2)

Figure 7 is the flow chart of the algorithm. The voltage needs tobe turned off if the robot rises too high, with a pressure lower thanPon. The voltage will be turned on when the robot drops too low,with a pressure higher than Poff.

The algorithm controls the robot to stay at a certain depth in ashort period. In the experiment, we place the robot at the bottomof the water tank with the initial depth of 25 cm. The robot takesabout 25 s (nearly four cycles of control) to reach the convergenceat the designated position (depth z¼ 15 cm.). The robot stays atthe designated depth, with the control signal oscillating (on andoff) at a period about 1.17 s. (Fig. 8).

In the experiment, the robot’s maximum variation of buoyantforce is 0.046 N when the voltage is 10 kV (DE membrane withtwo stacked VHB-4910). The mass of the robot is 0.45 kg. Themaximum acceleration is around 0.1 m/s�2. During one operationof the robot, the average speed of ascending reaches 3.5 cm/s.

4 Theoretical Analysis

We build up a mechanical model to analyze the swim-bladderrobot and to illustrate the floating mechanisms of robot driven by

Fig. 2 Actuation mechanism of the artificial swim bladder robot. (a) In the inflated state,the DE membrane was deformed by the pressure difference with no voltage applied. (b)In the actuated state, a high voltage is applied to the DE membrane. The electric fielddrives the ions in the surrounding water and electrons in the carbon grease. Positive andnegative charges accumulated on both sides of the DE membrane, inducing Maxwellstress and deforming the DE membrane. In the experiment with wired high voltage source,(c) the inflated balloon and chamber with a fixed total volume and buoyant force reachequilibrium at a certain depth under water. (d) The DE membrane was actuated by highvoltage. The volume and the buoyant force of the balloon increase. The robot floats up.

Journal of Applied Mechanics SEPTEMBER 2017, Vol. 84 / 091005-3

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 4 Buoyant force increment as a function of voltage, and the effects of initial inner pressure. (a) The rela-tion of buoyant force increment and voltage with various initial inner pressures. (b)–(g) The inflated balloonlikeshapes of the DE membrane with no voltage and different initial pressures of 106.0 kPa, 107.0 kPa, 107.5 kPa,108 kPa, 108.5 kPa, and 109.0 kPa.

Fig. 3 The onboard power source (Epod) and the method of control and communication. (a) Onboard high voltage source ispowered by a lithium–ion battery (3.7 V) with fly back topology to achieve small size, high voltage, and isolation. The circuitand battery were sealed in a plastic tube as the Epod. (b) The robot with an Epod assembled onboard. (c) The relation of con-trol, communication, and the actuation of the robot. (d) The data (collected from the pressure sensor) of pressure and voltagerelation with various voltages and membrane thickness.

091005-4 / Vol. 84, SEPTEMBER 2017 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

the actuation of DE membrane. The DE membrane is modeledusing the nonlinear theory of Suo [32]. Figure 9(a) illustrates amembrane of DE sandwiched between two compliant electrodes(the reference state). The membrane is subject to neither force norvoltage, and has a radius of R and a thickness of H. Under the

prestretched state, the membrane deforms with the radius of r andthe thickness of h. We simplify the DE balloon as a spherical shape.

Wsretch(k1,k2) is the free energy associated with the stretching ofthe elastomer membrane. The deformation of the membrane isinduced by the equal-biaxial Maxwell stresses with the magnitude

Fig. 5 The experimental results showing the relation of volume and buoyant force with various voltages andmembrane thickness, the effect of depth, and their corresponding numerical results from MATLAB simulation.(a) The relation of volume and voltage. (b) The numerical results of volume–voltage corresponding to theexperimental results in (a). (c) The relation of buoyant force and voltage. (d) The numerical results of buoyantforce increment–voltage corresponding to the experimental results in (c). (e) The relation of pressure andvoltage when the robot is anchored at the certain depth from 25.0 cm to 12.5 cm, respectively. (f) The numeri-cal results corresponding to the experimental results in (e).

Journal of Applied Mechanics SEPTEMBER 2017, Vol. 84 / 091005-5

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

of eE2, where e is the permittivity of the elastomer and E is theelectric field applied through the membrane [33–40]. We canadopt the Gent model [41] to describe the parameter Wsretch

Wsretch ¼ �lJlim

2log 1� k2

1 þ k22 þ k2

3 � 3

Jlim

!(3)

where l is the small-strain shear modulus of the elastomer andJlim is the material parameter related to the stretch limit.

In the absence of any applied loads, the state as shown inFig. 9(a) is considered as a state of reference. We define the coor-dinate’s origin at the center, and the z axis is normal to the plane.Any particle of radius R at the reference state may move to a placewith coordinates r and z at the current state when the membrane issubject to pressure p and voltage U (see Fig. 9(d)), where weassume that the membrane is of an axisymmetric shape.

With the boundary condition and initial value, we can get thevolume enclosed by the membrane

V ¼ðA

0

pr2 dz

dRdR (4)

where A is the radius of the membrane at the reference state(Fig. 9(a)), r and z are all the functions of R, and the equations arepartial differential equations, which lead to the insolubility of theanalytical results [42].

We simplify the analytical model and consider the DE balloonas a hemisphere. With the changing of air pressure or membrane

Fig. 6 The experimental data showing the relation of volume,voltage, and the control signal. (a) The relation of the volumeand the control signal received by the Epod. (b) The relation ofthe volume and the voltage generated from the external highvoltage source (Trek 610E). The dashed lines link the datapoints in (a) and (b), calibrating the voltage and the control sig-nal with the identical volume.

Fig. 7 The flow chart of the control algorithm. When the robotrises above the set point, the measured pressure will be lowerthan Pon, and the voltage will be turned off. When the robotsinks below the set point, the measured pressure will be higherthan Poff, the voltage will be turned off. Pon and Poff can be cal-culated from the experimental results.

Fig. 8 The fluctuation of the inner pressures during the opera-tion of control. (b) The fluctuation of inner pressure of the robotis large at the beginning of control. (a) The fluctuation of innerpressure of the robot is reduced and stabilized after control. Inthe period of the two dashed lines (80.88–87.77 s), there are sixcycles of pressure fluctuation.

Fig. 9 States of the membrane. (a) The reference state of mem-brane. (b) The prestretched state of membrane. (c) The inflatedstate of the membrane. (d) The actuated state of the membrane.(e) The schematic of the system with low applied voltage andsmall volume. (f) The schematic of the system with high appliedvoltage and large volume.

091005-6 / Vol. 84, SEPTEMBER 2017 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

tension, the radius of the DE balloon changes without changing itsspherical shape (Figs. 9(e) and 9(f)). When the robot reaches theequilibrium state, the relation between inner pressure in DE bal-loon, the external pressure of water, and the stress of the DE mem-brane can be expressed as

2pRsh � r0 � eUh

� �2 !

¼ pRs2 Pinner air � Pwaterð Þ (5)

where Rs is the radius of the membrane sphere, h is the thicknessof the membrane, Pinner air is the inner pressure of the robot, andPwater is the pressure of the water. r0 is the longitudinal and latitu-dinal stresses. Since the spherical DE membrane is thin, weneglect the stress normal to the membrane in the equation. Weassume that the material is incompressible, and the membrane isequal biaxially stretched with a spherical shape. The stretches ofthe membrane satisfy that

k1 ¼ k2 ¼ k; k3 ¼1

k1 � k2

¼ 1

k2(6)

k1 is the longitudinal stretch, k2 is the latitudinal stretch, and k3 isthe stretch in the thickness direction of the membrane. Then, r0

can be expressed as

r0 ¼ k1

@Wsretch k1; k2; k3ð Þ@k1

¼ l � k2 � k�4ð Þ1� 2k2 þ k�4 � 3ð Þ=Jlim

(7)

By comparing the experimental results with the model, weadjust the parameters to l¼ 45 kPa, Jlim¼ 270, and e¼ 4.16� 10�11 F/m [25].

Then, we simplify Eq. (5) to

l � k2 � k�4ð Þ1� 2k2 þ k�4 � 3ð Þ=Jlim

� eU � k2

H

� �2 !

� 2 H

k2

¼ Akffiffiffi2p

k0

Pinnerair � Pwaterð Þ (8)

where A is the radius of the acrylic chamber, and k0¼ 3 is the pre-stretch ratio from Figs. 9(a) and 9(b).

5 Control and Instability of the System

Assume that at a moment, the whole robot is in an equilibriumstate with no voltage applied, and the gravity equals to the buoy-ant force. The internal pressure in DE balloon, the external pres-sure of water, and the stress of the DE membrane satisfy that

2pRshr0 ¼ pR2s ðPinner air � PwaterÞ (9)

As the amount of the gas in the chamber is fixed, we assumethat the temperature does not change during the process. The pres-sure and volume of the robot under the initial and the currentstates satisfy the Ideal gas law

P0inner airV0 ¼ Pinner air V (10)

where P0inner air and V0 are the inner pressure and gas volume of thecurrent state, respectively. When there is a disturbance tothe robot causing a change of depth Dz (as shown in Fig. 10(a)),the water pressure variation due to the change of depth can beexpressed as

DP ¼ P0water � Pwater ¼ �qgDz (11)

As a consequence, the external pressure becomes smaller thanthe inner pressure. The DE membrane expands with a larger vol-ume, decreasing the inner pressure until the system reaches a newequilibrium state. Assume that stress of the DE membraneremains constant approximately

P0inner air � P0water ¼ Pinner air � Pwater ¼ 2pRshr0 (12)

DV ¼ Pinner air � P0inner air

P0inner air

V ¼ � DP

P0inner air

V (13)

where DV is the change of the volume.According to Newton second law

a ¼F0buoyancy � G� kv

M¼ � DP

P0inner air

� qgV

M� Kv (14)

where G is the gravity force, F0buoyancy is the buoyant force afterthe change, a is the acceleration of the robot, q is the water den-sity, g is local acceleration of gravity, V is the initial volume ofthe air chamber, M is the total mass of the robot, and k� vpresents the approximate fluidic resistance when the robot is atlow speed (v is the speed of robot and k is the fluidic resistancecoefficient). We define K¼ k/M as a constant. When a disturbancehappens with a small altitude range, P0inner air approximately equalsto Pinner air

a ¼ q2g2V

Pinner airMDz� Kv ¼ Cz� Kv set z0 ¼ 0ð Þ (15)

Fig. 10 The schematics of the robot during floating and sinking. (a) z is the vertical dis-placement, which can be calculated from the depths of the robot. (b) When the parametersof the system such as weight, initial pressure, and the thickness of DE membrane are fixed,the control method has a limit of operative depth. Only when the robot locates between theupper and lower limits, can the MCU control and stabilize the robot.

Journal of Applied Mechanics SEPTEMBER 2017, Vol. 84 / 091005-7

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

where C¼ (q2g2V)/(Pinner airM) is a constant, determined by therobot’s initial condition.

Let z¼ x1, v¼ x2, and a¼ _x2. The linear state space model ofthe system is

_x ¼ Ax ¼ 0 1

C �K

� �x (16)

The stability of the system can be investigated by the eigen-value equation as

Ds ¼ jsI� Aj ¼ s2 þ Ks� C ¼ 0 (17)

s1;2 ¼�K6

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiK2 þ 4Cp

2(18)

According to the first method of Lyapunov, as one of the eigen-values of s is positive, the system is unstable. Figure 11(a) showsthe block diagram of the unstable system.

In order to make the robot be able to stay at a designated depth,the system needs to be stable with a convergent output. We applyan on–off control to the system. Figure 11(b) shows the block dia-gram of the on–off control system. The sampling rate is 20 Hz.The water pressure is measured by the pressure sensor and trans-formed to the relative displacement of the robot, compared withcertain value according to the system’s condition and control volt-age applied on the DE membrane. The Maxwell stress induced bythe voltage relaxes the DE membrane and decreases r, and conse-quently changes the acceleration of the robot.

In order to analyze the performance of swim-bladder robot, wetake the robot condition (the curve in Fig. 5(c): thickness: 2 mm;initial pressure (above water): 108.0 kPa; initial depth: 25 cm) asexample, the maximum buoyant force is 0.046 N when the voltageis 10 kV. The mass of the robot is 0.45 kg. The maximum acceler-ation is around 0.1 m/s2. The robot’s operation range is the regionthat the MCU can control the system to converge at a certaindepth, which has a limitation. The upper bound of the controllabledepth is the depth that the robot starts to float up without applyingvoltage on DE membrane. The lower bound is the depth that therobot starts to sink down when the voltage already reaches themaximum value (10 kV in our experiment).

As shown in Fig. 10(b), we assume that the change of innerpressure is DP (voltage: 10 kV)

Dz ¼ DP

qg(19)

According to the experimental data, the controllable rangebetween upper bound and lower bound Dz is around 21.8 cm. Asshown in Fig. 3(d), when the thickness of the membrane increases,the maximum value of DP increases. So the controllable range islarger when the membrane is thicker. If the maximum voltage ofthe apparatus could be higher and the membrane could be thicker,the controllable range could be larger. However, the position ofupper bound and lower bound is decided by the balance weight.

6 Conclusions

In summary, we have developed an artificial swim-bladderrobot, which can be controlled by the actuation of an inflated DEmembrane. The air chamber, covered with the DE and filled withair, can provide buoyant force to maintain the robot balanced. Inaddition to the wireless mobility, the robot possesses other notableattributes including long endurance, low noise, and precise con-trol. The robot is driven by the DE membrane with applied vol-tages, which tune the pressure, total volume, and the buoyantforce of the robot. Stability analysis shows that the system isunstable, and the robot cannot maintain its depth without control.The simple on–off control can actuate the DE membrane withdesired shape to balance the buoyant force and maintain the posi-tion of the robot in designated depths. The thickness and the initialinflated shape of the DE membrane can affect the controllableoperating range of the robot. The robot achieves relatively highperformances in aspects such as buoyant force variation, speed,and acceleration. The robot can either function as a robotic systemitself or be attached on other robotic systems (robotic fish forexample) as a controllable floating module. The structural designand the control principles of this robot may guide the furtherdesign of aquatic soft robot and pressurized soft actuators.

Funding Data

� National Natural Science Foundation of China (11321202,11432012, 11572280, and U1613202).

� China Association for Science and Technology (Young EliteScientist Sponsorship Program).

References[1] Godaba, H., Li, J., Wang, Y., and Zhu, J., 2016, “A Soft Jellyfish Robot Driven

by a Dielectric Elastomer Actuator,” IEEE Rob. Autom. Lett., 1(2), pp.624–631.

[2] Rus, D., and Tolley, M. T., 2015, “Design, Fabrication and Control of SoftRobots,” Nature, 521(7553), pp. 467–475.

[3] Rogers, J. A., Someya, T., and Huang, Y., 2010, “Materials and Mechanics forStretchable Electronics,” Science, 327(5973), pp. 1603–1607.

[4] Trivedi, D., Rahn, C. D., Kier, W. M., and Walker, I. D., 2008, “Soft Robotics:Biological Inspiration State of the Art, and Future Research,” Appl. BionicsBiomech., 5(3), pp. 99–117.

[5] Kim, S., Laschi, C., and Trimmer, B., 2013, “Soft Robotics: A Bioinspired Evo-lution in Robotics,” Trends Biotechnol., 31(5), pp. 287–294.

Fig. 11 The block diagrams of the system. (a) The block dia-gram of the system without control. (b) The block diagram ofthe on–off control system. The sampling rate is 20 Hz.

091005-8 / Vol. 84, SEPTEMBER 2017 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

[6] Meng€uc, Y., Park, Y. L., Pei, H., Vogt, D., Aubin, P. M., Winchell, E., Fluke,L., Stirling, L., Wood, R. J., and Walsh, C. J., 2014, “Wearable Soft SensingSuit for Human Gait Measurement,” Int. J. Rob. Res., 33(14), pp. 1748–1764.

[7] Su, Z., Yu, J., Tan, M., and Zhang, J., 2014, “Implementing Flexible and FastTurning Maneuvers of a Multijoint Robotic Fish,” IEEE/ASME Trans. Mecha-tronics, 19(1), pp. 329–338.

[8] Yamakita, M., Kamamichi, N., Kozuki, T., and Asaka, K., 2005, “A Snake-Like Swimming Robot Using IPMC Actuator and Verification of DopingEffect,” IEEE/RSJ International Conference on Intelligent Robots and Systems(IROS), Edmonton, AB, Canada, pp. 2035–2040.

[9] Villanueva, A., Smith, C., and Priya, S., 2011, “A Biomimetic Robotic Jellyfish(Robojelly) Actuated by Shape Memory Alloy Composite Actuators,” Bioinspi-ration Biomimetics, 6(3), p. 036004.

[10] Kopman, V., and Porfiri, M., 2013, “Design, Modeling, and Characterization ofa Miniature Robotic Fish for Research and Education in Biomimetics and Bio-inspiration,” IEEE/ASME Trans. Mechatronics, 18(2), pp. 471–483.

[11] Xiao, P., Yi, N., Zhang, T., Huang, Y., Chang, H., Yang, Y., Zhou, Y., andChen, Y., 2016, “Construction of a Fish-Like Robot Based on High PerformanceGraphene/Pvdf Bimorph Actuation Materials,” Adv. Sci., 3(6), p. 1500438.

[12] Wang, Z., Hang, G., Li, J., Wang, Y., and Xiao, K., 2008, “A Micro-Robot FishWith Embedded SMA Wire Actuated Flexible Biomimetic Fin,” Sens. Actua-tors A, 144(2), pp. 354–360.

[13] Tolley, M. T., Shepherd, R. F., Mosadegh, B., Galloway, K. C., Wehner, M.,Karpelson, M., Wood, R., and Whitesides, G., 2014, “A Resilient, UntetheredSoft Robot,” Soft Rob., 1(3), pp. 213–223.

[14] Shepherd, R. F., Ilievski, F., Choi, W., Morin, S. A., Stokes, A. A., Mazzeo, A.D., Chen, X., Wang, M., and Whitesides, G. M., 2011, “Multigait Soft Robot,”Proc. Natl. Acad. Sci. U.S.A., 108(51), p. 20400.

[15] Morin, S. A., and Whitesides, G. M., 2012, “Camouflage and Display for SoftMachines,” Science, 337(6096), pp. 828–832.

[16] Seok, S., Onal, C. D., Cho, K. J., and Wood, R. J., 2013, “Meshworm: A Peri-staltic Soft Robot With Antagonistic Nickel Titanium Coil Actuators,” IEEE/ASME Trans. Mechatronics, 18(5), pp. 1485–1497.

[17] Graule, M. A., Chirarattananon, P., Fuller, S. B., Jafferis, N. T., Ma, K. Y.,Spenko, M., Kornbluh, R., and Wood, R. J., 2016, “Perching and Takeoff of aRobotic Insect on Overhangs Using Switchable Electrostatic Adhesion,” Sci-ence, 352(6288), pp. 978–982.

[18] Park, Y. L., Chen, B. R., P�erezarancibia, N. O., Young, D., Stirling, L., Wood,R. J., Goldfield, E. C., and Nagpal, R., 2014, “Design and Control of a Bio-Inspired Soft Wearable Robotic Device for Ankle-Foot Rehabilitation,” Bioins-piration Biomimetics, 9(1), p. 016007.

[19] Polygerinos, P., Wang, Z., Galloway, K. C., Wood, R. J., and Walsh, C. J.,2015, “Soft Robotic Glove for Combined Assistance and At-Home Rehabil-itation,” Rob. Auton. Syst., 73, pp. 135–143.

[20] Ilievski, F., Mazzeo, A. D., Shepherd, R. F., Chen, X., and Whitesides, G. M.,2011, “Soft Robotics for Chemists,” Angew. Chem., 123(8), pp. 1930–1935.

[21] Park, S. J., Gazzola, M., Park, K. S., Park, S., Di Santo, V., Blevins, E. L., Lind,J. U., Campbell, P. H., Dauth, S., Capulli, A. K., Pasqualini, F. S., Ahn, S.,Cho, A., Yuan, H., Maoz, B. M., Vijaykumar, R., Choi., J. W., Deisseroth, K.,Lauder, G. V., Mahadevan, L., and Parker, K. K., 2016, “Phototactic Guidanceof a Tissue-Engineered Soft-Robotic Ray,” Science, 353(6295), pp. 158–162.

[22] Nawroth, J. C., Lee, H., Feinberg, A. W., Ripplinger, C. M., Mccain, M. L.,Grosberg, A., Dabiri, J. O., and Parker, K. K., 2012, “A Tissue-Engineered Jel-lyfish With Biomimetic Propulsion,” Nat. Biotechnol., 30(8), pp. 792–797.

[23] Koh, J. S., Yang, E., Jung, G. P., Jung, S. P., Son, J. H., Lee, S. I., et al., 2015,“Biomechanics. Jumping on Water: Surface Tension-Dominated Jumping ofWater Striders and Robotic Insects,” Science, 349(6247), pp. 517–521.

[24] Laschi, C., Cianchetti, M., Mazzolai, B., Margheri, L., Follador, M., and Dario,P., 2012, “Soft Robot Arm Inspired by the Octopus,” Adv. Rob., 26(7), pp.709–727.

[25] Lin, H. T., Leisk, G. G., and Trimmer, B., 2011, “Goqbot: A Caterpillar-Inspired Soft-Bodied Rolling Robot,” Bioinspiration Biomimetics, 6(2),p. 026007.

[26] Marchese, A. D., Onal, C. D., and Rus, D., 2014, “Autonomous Soft RoboticFish Capable of Escape Maneuvers Using Fluidic Elastomer Actuators,” SoftRob., 1(1), pp. 75–87.

[27] Wehner, M., Truby, R. L., Fitzgerald, D. J., Mosadegh, B., Whitesides, G. M.,Lewis, J. A., and Wood, R. J., 2016, “An Integrated Design and FabricationStrategy for Entirely Soft, Autonomous Robots,” Nature, 536(7617), pp.451–455.

[28] Li, T., Li, G., Liang, Y., Cheng, T., Dai, J., Yang, X., Liu, B., Zeng, Z., Huang,Z., Luo, Y., Xie, T., and Yang, W., 2017, “Soft Electro-Ionic Fish With Auton-omous Fast Moving,” Sci. Adv., 3(4), p. e1602045.

[29] Anderson, I. A., Kelch, M., Sun, S., Jowers, C., Xu, D., and Murray, M. M.,2013, “Artificial Muscle Actuators for a Robotic Fish,” Biomimetic and Biohy-brid Systems, Vol. 8064, Springer, Berlin, pp. 350–352.

[30] Jordi, C., Michel, S., D€urager, C., Bormann, A., Gebhardt, C., and Kovacs, G.,2010, “Large Planar Dielectric Elastomer Actuators for Fish-Like Propulsion ofan Airship,” Proc. SPIE, 7642, pp. 579–595.

[31] Carpi, F., Carpi, F., De Rossi, D., Kornbluh, R., Pelrine, R., and Sommer-Larsen, P., 2008, Dielectric Elastomers as Electromechanical Transducers(Fundamentals, Materials, Devices, Models and Applications of an EmergingElectroactive Polymer Technology), Elsevier Science, Oxford, UK.

[32] Suo, Z., 2010, “Theory of Dielectric Elastomers,” Acta Mech. Solidia Sin.,23(6), pp. 549–578.

[33] Wissler, M., and Mazza, E., 2005, “Modeling of a Pre-Strained Circular Actua-tor Made of Dielectric Elastomers,” Sens. Actuators A, 120(1), pp. 184–192.

[34] Wissler, M., and Mazza, E., 2005, “Modeling and Simulation of DielectricElastomer Actuators,” Smart Mater. Struct., 14(6), pp. 1396–1402.

[35] Goulbourne, N., Mockensturm, E., and Frecker, M., 2005, “A Nonlinear Model forDielectric Elastomer Membranes,” ASME J. Appl. Mech., 72(6), pp. 899–906.

[36] Goulbourne, N. C., Mockensturm, E. M., and Frecker, M. I., 2007, “Electro-Elastomers: Large Deformation Analysis of Silicone Membranes,” Int. J. SolidsStruct., 44(9), pp. 2609–2626.

[37] Plante, J. S., and Dubowsky, S., 2006, “Large-Scale Failure Modes of DielectricElastomer Actuators,” Int. J. Solids Struct., 43(25), pp. 7727–7751.

[38] Plante, J. S., and Dubowsky, S., 2007, “On the Performance Mechanisms ofDielectric Elastomer Actuators,” Sens. Actuators A, 137(1), pp. 96–109.

[39] Zhao, X., and Suo, Z., 2008, “Electrostriction in Elastic Dielectrics UndergoingLarge Deformation,” J. Appl. Phys., 104(12), p. 123530.

[40] Li, B., Chen, H., Qiang, J., and Zhou, J., 2012, “A Model for Conditional Polar-ization of the Actuation Enhancement of a Dielectric Elastomer,” Soft Matter,8(2), pp. 311–317.

[41] Gent, A. N., 1996, “A New Constitutive Relation for Rubber,” Rubber Chem.Technol., 69(1), pp. 59–61.

[42] Li, T., Keplinger, C., Baumgartner, R., Bauer, S., Yang, W., and Suo, Z., 2013,“Giant Voltage-Induced Deformation in Dielectric Elastomers Near the Vergeof Snap-Through Instability,” J. Mech. Phys. Solids, 61(2), pp. 611–628.

Journal of Applied Mechanics SEPTEMBER 2017, Vol. 84 / 091005-9

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 09/06/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use