This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE/ASME TRANSACTIONS ON MECHATRONICS 1

Bistable Antagonistic Dielectric Elastomer Actuators for Binary Robotics and Mechatronics
Patrick Chouinard and Jean-S ebastien Plante
AbstractBinary systems can lead to simple and efcient robotic and mechatronic systems since such systems use a large number of simple bistable actuators to affect its state. Dielectric elastomer actuators (DEAs) are prime candidates for use in binary systems since they are simple, low cost, and lightweight. However, previously proposed bistable DEAs (ip-op) have relatively low volumetric energy density that limits their use in practical devices. This paper investigates the potential of improving the energy density of bistable designs by employing DEAs in compact antagonistic congurations. To do so, two antagonistic congurations (linear and rotating) are designed and studied using an experimentally validated BergstromBoyce viscoelastic material model. The proposed antagonistic congurations show up to 10 higher volumetric energy densities than ip-op designs. This represents a signicant advantage for DEA reliability, since, based on volumetric energy density, antagonist actuators require the manufacturing of signicantly less lm layers than ip-op designs. This study also reveals that, in the design of antagonistic DEAs, limiting the polymer lms actuation stretch minimizes viscoelastic losses and allows higher actuation speeds and power outputs for a given actuator stroke and size. Index TermsAntagonistic, bistable, dielectric elastomer actuator (DEA), viscoelastic model.

Fig. 1.

DEA working principle.

I. INTRODUCTION

INARY actuation is similar to digital electronics in the way that a large number of binary actuators ip between two discrete states [1]. Discrete positioning is highly repeatable and renders binary devices potentially simpler and more economical than current analog systems since they do not require low-end feedback control [2], [3]. Binary devices also have higher reliability because they possess redundant actuators; if one fails, the device remains operational. Binary devices require a large number of actuators to match the resolution and precision of current continuous devices [2]. Current actuator technologies (e.g., solenoids and pneumatic actuators) are too bulky and expensive to be used in large numbers in binary devices [4]. Smart materials offer an interesting alternative because of their potential for high volumetric energy densities and for integration in large number into a structure. Some smart material actuators showed promising perfor-

Manuscript received October 26, 2010; revised February 15, 2011 and March 2, 2011; accepted March 6, 2011. Recommended by Technical Editor G. Alici. The authors are with the Universit e de Sherbrooke, Sherbrooke, QC J1K 2R1, Canada (e-mail: patrick.chouinard@usherbrooke.ca; Jean-sebastien.plante@ usherbrooke.ca). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TMECH.2011.2135862

mances but possess distinct disadvantages that make them inadequate for use in practical binary devices. For example, shape memory alloys have low efciencies, ionic polymers have limited force output [5], and piezoelectric ceramics have very low strains [4]. Dielectric elastomer actuators (DEAs) are promising for binary robotics because of their simplicity, low cost, and lightweight. Among their good all around performances, they have the potential for large strains (more than 100%) and high theoretical energy to weight ratio (up to 34 J/g) [6]. DEAs offer high versatility since they can take multiple forms, such as stack actuators, minimum energy structures, diamond actuators, etc. [7]. DEAs studied in this paper are composed of a prestretched polymer lms held in between rigid frames, where compliant electrodes are applied on each side of the lm. As seen in Fig. 1, when voltage is applied to the electrodes, Maxwell stresses compress the polymer lm that causes it to deform [8]. Acrylic and silicone lms are generally used in DEAs because of their high dielectric constant, breakdown strength, and elongation at break [9]. Acrylic lms, such as VHB 4905 from 3M, are commonly used in actuators where high energy density and large strains (200%) are needed, but the materials strong viscoelastic nature limits actuators bandwidth to 3040 Hz [10]. Silicone lms with low viscoelastic effects are generally required for fast acting DEAs. However, their energy density and strain output (20%) are signicantly low compared to acrylic lms [11]. DEAs are susceptible to pull-in failure and current leakage when voltage is continuously applied to the electrodes [12]. In contrast, bistable DEAs have greater reliability and efciency since they do not require electrical energy to maintain a stable state. Therefore, voltage is only applied when the actuator shifts between its stable positions. In past research, ratcheting mechanisms and buckling beam bistable elements were coupled to DEAs to achieve bistability. The MiniMight concept used a single diamond-shape DEA to drive a ratcheting mechanism [13]. This concept was capable of high volumetric energy densities since it used a single planar actuator. However, the ratcheting mechanism was complex and

1083-4435/$26.00 2011 IEEE

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE/ASME TRANSACTIONS ON MECHATRONICS

Fig. 2. Flip-op bistable actuators using (a) diamond shape [13] and (b) cone shape DEAs [14].

DEAs made from acrylic lms experience signicant performance losses due to the viscoelastic nature of the polymer [16]. Viscous losses have even more impact on antagonistic actuators since the active actuation cell must drive the opposing inactive one. DEAs made from low viscoelasticity materials, such as silicone, is generally a viable solution to minimize viscoelastic losses. However, antagonistic actuators made from silicone lms have low strokes and work output due to the very limited strains of silicone lms. For example, preliminary experiments showed that antagonistic actuators using silicone lms (CF192186) have a work output 10 inferior to actuators using acrylic lms (VHB 4905). Therefore, only antagonistic congurations using acrylic lms are studied in this paper. A. Approach and Results The main objective of this paper is to verify the ability of antagonistic DEAs to provide simpler and higher energy dense bistable actuators than ip-op devices. For this purpose, a linear and a rotating conguration are designed and studied using acrylic lms (VHB 4905). Design and performance analysis are carried out using an experimentally validated 1-D Bergstrom Boyce analytical model. The performances of the antagonistic actuators are then benchmarked in terms of energy and power density against existing ip-op designs. This study shows that antagonistic actuators are capable of 10 the volumetric energy density of ip-op devices. This study also reveals a design tradeoff between actuator shifting speed, energy density, and power density through tuning the polymer lms actuation stretch. It is shown that by lowering the materials actuation stretch, although resulting in a slight decrease in energy density, rotating antagonistic actuators can have up to 10 more power density output than rotating actuators that are designed solely for high energy density.

Fig. 3. (a) 3-D exploded model view of the linear conguration. (b) Schematic view of the linear concept. (c) Forces acting on the actuators center piece.

II. ANTAGONISTIC BISTABLE CONCEPTS would have presented reliability and manufacturing limitations when used in large numbers in binary devices. Flip-op devices (see Fig. 2) combined two DEAs positioned face-to-face to drive a buckling beam bistable element. These designs had a considerable volume that limited the maximum work developed by the actuator for a given size [14]. Bistable DEAs with the simplicity of ip-op devices but with high volumetric energy densities are, therefore, required so that DEAs can be used in practical binary devices. Antagonistic congurations refer to DEAs with opposing actuation cells (a.k.a. electrodes) dened on the same lm [see Fig. 3(a)]. The system has the particularity that the motion of an actuation cell is contended by the other one. Antagonistic DEAs could provide higher energy densities than ip-op devices due to lower dead volume enabled by a simpler design and a planar construction well suited for laminate assembly. The theoretical principle of using antagonistic actuators to achieve bistability has been exposed by Wingert et al. [15] but has not been conceptualized in practical actuator designs. A. Linear Conguration The linear conguration in Fig. 3(a) uses a carbon leaf spring assembly, acting as a bistable element, sandwiched in between layers of polymer lm. Fig. 3(b) illustrates a schematic view of the linear conguration, where the carbon leaf springs are schematized by linear springs. Fig. 3(c) shows the forces of the lms Flm and springs Fspring acting on the actuators center piece during a shift. Bistability is achieved when the springs are tuned so that their force response is higher than the force of the lm when the center piece is up against a stopper and no voltage is applied. When voltage is applied to the proper actuation cell, the resulting force of the lm becomes stronger than the force of the springs causing the center piece to shift toward its other stable position. The planar conguration has a minimal dead volume that maximizes energy density and that can be used in closely packed arrays. This conguration could be useful in applications such as microscale robots [17], Braille displays, and active grasping surfaces.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
CHOUINARD AND PLANTE: BISTABLE ANTAGONISTIC DEAs FOR BINARY ROBOTICS AND MECHATRONICS 3

Fig. 6.

Micromechanism of the BergstromBoyce model.

III. ANALYTICAL MODEL The design of antagonistic congurations is based upon a continuum mechanics model that accounts for the large strains, nonlinear properties, and viscoelastic nature of the polymer. A 1-D BergstromBoyce constitutive viscoelastic model is used as it has shown good agreement with experimental results of diamond-shaped acrylic actuators undergoing large stretches at rates varying by three orders of magnitude [16]. Although essential for the future of DEAs, nite element models that reliably include the time-dependent behavior of complex actuator shapes over large stretches and widely varying stretch rates have yet to be developed [18]. Hence, a robust 1-D viscoelastic model is better suited to study the dynamics of antagonistic actuators. A. BergstromBoyce Viscoelastic Model The BergstromBoyce model is inspired by a polymer chain micromechanism where viscoelasticity originates from the timedependent motion of the polymer chains relative to each other. This micromechanism, shown in Fig. 6, is composed of a purely elastic polymer chain network (Network A) superimposed to a time-dependent network (Network B) [19]. As seen in Fig. 6, the materials total deformation gradient Ftot is equal to the deformation gradient of both networks (FA = FB = Ftot ). Using the KronerLee decomposition, Ftot is decomposed into the deformation gradient of the elastic (FE B ) and time-dependent element of network B (FP B ):
E Ftot = FP B FB .

Fig. 4. (a) 3-D Rotating conguration model view. (b) Schematic view of the rotating conguration. (c) Plot of the torque applied about the hinge pivot.

Fig. 5. (a) Quadristable rotating concept. The four stable states are dened with a cross-shaped gating pattern that guides the ball joint in two orthogonal directions. (b) Four stable positions of the quadristable rotating design. Voltage is applied on the actuation cells colored in white. (c) Robotic positioning system.

B. Rotating Conguration The rotating conguration in Fig. 4(a) uses layers of lm that are prestretched in a planar state and then pushed out-of-plane by raising a hinge mechanism. This conguration has the advantage of being capable of intrinsic bistability, meaning that it does not necessitate the use of external springs to gain bistability. Bistability is achieved geometrically with an over-the-center mechanism. Such intrinsically bistable actuators require fewer components and are, consequently, much simpler and reliable than designs that use springs. Further, intrinsic bistability is not a function of the number of lm layers, making the design easily scalable for various force outputs. Fig. 4(b) shows a schematic view of the rotating conguration and Fig. 4(c) the torque generated by the forces of the lm. As seen in Fig. 4(c), the torque generated by the lm Tlm holds the top hinge up against the stopper when no voltage is applied. When voltage is applied to the proper actuation cell, Tlm orientation inverses and causes the actuator to shift toward the adjacent stopper. Ultimately, the development of the rotating conguration could lead to the development of multistable actuators that could be implemented in highly simple binary robotic and mechatronic devices such as a compliant manipulator arm (see Fig. 5). Potential applications for these systems are walking robots, highprecision manipulators, cost-effective consumer products, and multiple DoF haptic devices.

(1)

This model stipulates that the rate of the deformation gradient P ) is calculated using the ow of the time-dependent element (F B rate of the networks stress deviatoric tensor [19]. Cauchy stresses tot are evaluated by summing the stresses in both networks. The stress of the elastic elements is evaluated using a rst-order hyperelastic incompressible Ogden model expressed in terms of the material principal stretches taken in the three main directions (i = 1, 2, 3) [20]. The materials stretches are expressed in three states: prestretch pre ,i , actuation act , i , and total tot ,i , such as tot ,i = act ,i pre ,i = ltot ,i lpre ,i ltot ,i = lpre ,i lfree ,i lfree ,i (2)

where l is the length of the polymer lm at a given state (total, prestretch or free) and direction. Note that the direction i = 1 refers to the direction parallel to the displacement of the actuator

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4 IEEE/ASME TRANSACTIONS ON MECHATRONICS

TABLE I IMPOSED STROKES AND SPEEDS USED FOR THE MATERIAL CHARACTERIZATION OF THE ANTAGONISTIC CONFIGURATIONS

Fig. 8. Curve t of a rotating conguration (a) shifting at ram speed of 1 mm/s (1.5 /s) and (b) ensuing force relaxation after the shift.

Fig. 7. Experimental setup of (a) linear conguration and (b) rotating conguration (side view).

Fig. 9. Curve t of a rotating conguration (a) shifting at ram speed of 10 mm/s (15 /s) and (b) ensuing force relaxation after the shift.

and i = 3 to the direction parallel to the applied Maxwell pressure (see Fig. 1). The force generated by the electrostatic eld is modeled by an equivalent Maxwell pressure expressed by [8] PM ax = die 0 V u
2

(3)

where 0 is the dielectric constant of permittivity, die the lms dielectric constant, V the voltage applied to the actuation cell, and u the lms thickness. The model applies the equivalent Maxwell pressure to Network A and evaluates its value at every time step with regard to the instantaneous thickness of the lm. The analytical model is implemented and numerically solved using a MATLAB/Simulink software. B. Material Characterization BergstromBoyce material constants are tted to each actuator conguration using their experimental force curves while shifting at various strokes and speeds, including the postshift stress relaxation. Table I indicates the stroke and speed range used to characterize the linear and rotating congurations. For optimal precision, material characterization is realized using actuator geometries that are similar to the ones of the nal prototypes [20]. Uniaxial force curves are acquired using a TA-XT plus Texture Analyser with the experimental setup shown in Fig. 7. The ram of the Texture Analyser cycles the actuators center piece from one stopper position to the other while the actuators frame is xed to the base of the machine. Force output is acquired by a load cell that is xed to the machines ram. For the rotating conguration, actuators torque output is deduced from the acquired force and the geometry of the lever arms. Tested DEAs are built out of laminates of three layers of VHB 4905. Each acrylic lm layers is prestretched asymmetrically by pre , 1 pre , 2 = 3 4. Film prestretch values are

selected to be as high as possible without reaching the materials tearing point in order to maximize actuators stroke and minimize viscous losses [16], [21]. Experiments performed on the linear conguration with different prestretch values conrmed the approach. The prestretch, electrode deposition and frame assembly techniques that are used to build the prototypes are based on manual but reliable manufacturing methods detailed in [22]. Electrodes are made from a mixture of carbon black and toluene and are airbrushed onto the lm [23]. An assembly jig is used to position accurately the center piece and the frame on the acrylic lm. The lms dielectric constant (die = 3.2) is validated by curve tting (3) on experimental force curves acquired when voltage is applied to an actuation cell. Even though research has shown the polymers dielectric constant to vary with its total stretch and actuation frequencies, it is assumed constant here as its slight variations have limited effects on actuator performances [21]. IV. EXPERIMENTAL RESULTS As shown in Figs. 8 and 9, the analytical model correlates well with the DEAs experimental force curves for constant shift speeds (as imposed by TA-XT ram speed) varying by one order of magnitude. Optimal ts between analytical and experimental curves are obtained when different material constants are attributed to specic actuator geometry. This is mostly due to the 1-D model that neglects the force response of the inactive portions of lm (no electrode), which varies from one actuator geometry to the other. Additionally, the true lm deformed shape of the rotating conguration has an inward curvature on the actuation cells while it is considered to be straight in the analytical model. Also, inherent variability of the acrylic lms properties from roll to roll adds to the uctuations of material constants. Nonetheless, through proper model calibration, the force response of the DEA can be

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
CHOUINARD AND PLANTE: BISTABLE ANTAGONISTIC DEAs FOR BINARY ROBOTICS AND MECHATRONICS 5

predicted accurately even though the full 3-D force response of the lm is neglected.

V. DESIGN STUDY This section details the design of antagonistic actuators. First, the mechanics of linear and rotating antagonistic actuators is explained with emphasis on the performance losses due to the viscoelastic nature of the lm. Second, a nondimensional analysis is performed with the objective of selecting the geometric parameters that maximize the rotating congurations energy density. Finally, the performances of the linear and rotating congurations that have an optimized geometry are analyzed using predictions from the experimentally calibrated analytical model. The following considerations are taken into account: 1) all actuator designs are evaluated using a voltage equivalent to 70% of the breakdown strength of the lm published in [24] and with a center pieces length (L) of 100 mm; 2) the minimal total stretch of the actuation cells is set to tot , 1 = 1.1 to prevent the lm from buckling; 3) the maximal stroke of the rotating conguration is set to 40 to prevent the polymer lm from colliding with the center hinge; and 4) the distance between ends of the center piece and the frame are set to avoid the materials tearing point.

Fig. 10. (a) Analytical shifting curves and proposed spring curve of a linear DEA shifting at very low speeds. (b) Analytical shifting curves at 0.1 and 0.9 mm/s.

A. Linear Conguration The design of antagonistic bistable DEAs is based on their shifting and force curves. Fig. 10(a) illustrates the shifting and force curves of the linear conguration undergoing a 8 mm stroke at very low speeds, where viscoelastic forces are negligible (below equilibrium speed). The shifting curve refers to the lms force response when an actuation cell is activated and the force curve, when neither actuation cell is activated. As seen in Fig. 10(a), the output force of the linear conguration is Fon when the actuators is shifting (V = 10 kV) and Fo when in a stable position (V = 0 kV). To maximize the actuator output force, the springs force curve must be designed to t halfway in between the actuators shifting and force curve, so that Fo = Fon at the maximum stroke. Consequently, the actuators lowest output force is found at the end of an actuator shift. When the actuator is forced to shift at higher speeds, opposing viscoelastic forces build into the actuator and make the shifting curves come closer to each other [see Fig. 10(b)]. This causes further restrictions in the spring design space and lowers the output force of the actuator. Ultimately, no spring force curve can be tted in between the actuators shifting curves as they will cross each other, thereby dening a maximum actuation speed. The variation of the shifting curves with actuation speeds also indicates that, for optimal performances (Fo = Fon ), the leaf springs must be tted according to the desired operational speed of the actuator.

Fig. 11. (a) Analytical shifting and torque curves of the rotating conguration shifting at very low speeds. (b) Analytical shifting curves at 1 /s and 25 /s.

B. Rotating Conguration Fig. 11(a) traces intrinsically bistable rotating congurations shifting and torque curves undergoing a 40 stroke at very low speed. In this case, the actuators torque output is maximal when the actuators geometry is tuned so that To = Ton around the abscissa. Similar to the linear conguration, when actuation speed increases, the actuators torque output decreases since the shifting curves approach the abscissa [see Fig. 11(b)]. The maximal actuation speed of rotating congurations occurs when the shifting curves reach the abscissa given that the actuators torque output is null.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6 IEEE/ASME TRANSACTIONS ON MECHATRONICS

TABLE II CHOSEN GEOMETRIC PARAMETERS

Fig. 12.

Geometric parameters of the rotating conguration.

Fig. 14.

Prototype of (a) linear and of (b) rotating conguration .

However, lower actuation stretches result in lower stretch rates and, thus, lower viscous blockage. Hence, Conguration is selected to illustrate a design tradeoff between shifting speed, energy density, and power density.
Fig. 13. (a) Energy density potential (J/m3 ) and (b) actuation stretch of the rotating congurations at stroke of 40 at m/M = 0.2.

D. Actuators Performance Analysis Model predicted shifting and force/torque curves of the antagonistic DEAs are used to study the chosen concepts work and power outputs for a wide range of actuation speeds. To ensure the accuracy of the results, the analytical model is calibrated with experimental curves of the linear and rotating congurations ( and ) shown in Fig. 14. When calibrated to specic actuator geometry, the developed model is sufciently accurate to predict the dynamic behavior of antagonistic DEAs. Fig. 15(a) shows that the analytical and experimental actuation speeds vary similarly throughout the shift. Furthermore, Fig. 15(b) illustrates that the average analytical actuation speed corroborates relatively well to the experimental results in the actuators operating stroke range (20 to 40 ). However, at lower strokes, the average analytical shifting speed is 3 higher than experimental shifting speed. The proposed BergstromBoyce model accuracy at low strokes could be improved either through better calibration in this stroke range or through a more accurate mathematical formulation of the ow rate function. The work and power output of the linear and rotating ( and ) congurations are plotted over a wide range of actuation speeds using the experimentally calibrated BergstromBoyce model (see Figs. 16 and 17). As seen in Figs. 10(b) and 11(b), the force/torque output of the actuators is not constant throughout the shift. The energy and power density of the actuators are, therefore, evaluated using the minimal force/torque output of the actuators during a shift. The results presented in Figs. 16 and 17 apply for a single lm layer. The actuators work and power output are linearly

C. Nondimensional Analysis As the rotating congurations geometry offers great design exibility, a nondimensional analysis is carried out to select sets of geometric parameters that maximize energy density. In this analysis, shifting speed is supposed to be very low, where viscoelastic losses can be neglected. This corresponds to the highest possible work output of the actuator. Mathematically, the viscoelastic forces are removed by only considering Network A of the BergstromBoyce model (see Fig. 6). The geometric parameters used in this study are m/M, h2 /M, and z/M and are shown in Fig. 12. Fig. 13(a) illustrates the actuators energy density potential in terms of the nondimensional parameters z/M and h2 /M. In all cases, the actuators energy density increases for decreasing m/M; hence, this parameter is set to a minimal value of 0.2 to provide an aspect ratio that satises manufacturing constraints. As shown in the previous section, for maximal energy output, the rotating conguration requires To and Ton to be equal around the abscissa. This condition is represented by the dash line in Fig. 13(a). Fig. 13(b) illustrates the corresponding actuation stretch act , 1 showing that high energy densities occur at high actuation stretches. Based on these results, two potential sets of geometric parameters leading to high energy densities are identied in Fig. 13 as and . The geometric parameters of the selected congurations are summarized in Table II. Conguration represents the highest value of energy density potential and has consequently maximal actuation stretch act , 1 .

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
CHOUINARD AND PLANTE: BISTABLE ANTAGONISTIC DEAs FOR BINARY ROBOTICS AND MECHATRONICS 7

Fig. 18. Holding time of the (a) linear conguration, (b) Conguration , and (c) Conguration at 75% and 100% output force. Fig. 15. Conguration As analytical and experimental (a) top hinge position during a no load shift and (b) average shifting speed at different strokes. All shifting speeds are plotted using a supplied voltage of 8500 V.

Fig. 16. Output work in terms of actuation speeds of the (a) linear conguration, (b) Conguration at strokes of: 20 , 30 , 40 , and (c) Conguration at stroke of 40 .

Fig. 11). Given that DEAs operating at high speed exhibit lower output force, less viscous forces are built into the actuator; thus, less time is needed to gain bistability. Consequently, holding time decreases considerably when it is evaluated for an output force corresponding to 75% of the maximal output force [see Fig. 18(b) and (c)]. Holding time can have a signicant impact on reliability since applying high voltages for extended periods of time increases the risk of pull-in and dielectric strength failures [12]. To increase reliability, actuators must then be designed to maximize actuation speed to reduce holding time. However, fast response antagonistic actuators with low holding time would experience a reduced output force resulting from high-speed operation, imposing another design tradeoff. VI. DISCUSSION This section discusses the advantages of using antagonistic actuators over ip-op actuators for practical binary applications. To do so, the performances of the linear and rotating congurations (see Table III) are benchmarked according to the published performances of ip-op actuators [12]. The four parameters used in this study are the max average speed, energy per volume of lm, energy density, and power density. The max average speed is evaluated using the time required by the actuator to complete a full stroke shift under no load. Figs. 16 and 17 are used to set the energy and power output of the actuators. The prototypes physical dimensions are used to evaluate the actuators total and lm volume. Energy and power densities are estimated by dividing the actuators energy and power output by the actuators volume. Energy per volume of lm: The antagonistic congurations work output per volume of lm is inferior to ip-op devices as half of its active material is used to provide output work while the other half dissipates energy through viscous losses. Energy density: The work output with regard to the total actuators volume is similar between all antagonistic congurations and is at least 10 superior to ip-op devices. This is due to the compact design of antagonistic actuators that combines two actuation cells on individual lm layers and frame. Flip-op devices have large dead volumes since they contain two separate DEAs and a bistable element. The higher energy density of antagonistic actuators has a signicant advantage on reliability. Adding lm layers to DEAs linearly increases work output without considerably affecting actuator volume, and thus, ip-op devices could match the energy density of antagonistic congurations by adding 10

Fig. 17. Output power in terms of actuation speeds of the (a) linear conguration, (b) Conguration at different strokes of 20 , 30 , 40 , and (c) Conguration at stroke of 40 .

scalable with the number of lm layers regardless of actuation speed. The effect of actuator stroke on performance can be seen in Figs. 16(b) and 17(b). Fig. 17(b) shows that performing lower strokes leads to greater actuation speed and thus power output. This is due to the acrylic lm that increasingly builds up viscous blockage as actuation stretch increases. This implies a design tradeoff since, as seen in Fig. 16(b), lower actuation stretches yield lower work output. This tradeoff is also witnessed when comparing Congurations and . For a stroke of 40 , since Conguration has a higher actuation stretch than Conguration , it is capable of greater work output (3) but lower power output (2). E. Holding Time Voltage needs to be continuously applied for an extended period of time after the shift (a.k.a. holding time) so that antagonistic bistable DEAs reach bistability and their maximal output force. Holding time corresponds to the time needed to relax the viscoelastic forces of the material to a given level. As shown in Fig. 18, holding time is highest at low shifting speeds and decreases at higher actuation speeds. This is explained by the exponential decay of the viscoelastic forces (see

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8 IEEE/ASME TRANSACTIONS ON MECHATRONICS

TABLE III EXPERIMENTAL PERFORMANCES OF BISTABLE DEA CONCEPTS

more lm layers (per actuator). However, for a same energy density and manufacturing process, antagonistic actuators can be more reliable since they use fewer lm layers and thus have less chances of including a defective layer. This is an important consideration given that current DEAs manufacturing techniques present reliability issues [11]. Power density: By using acrylic lms in an actuator geometry that minimizes the lms actuation stretch, the materials viscoelastic properties become less dominant and greater actuation speed, thus power output, is achievable. This is conrmed with Conguration that is capable of 4 the power density of ip-op devices and 10 the power density of Conguration . All other bistable concepts use acrylic lm in a conguration that maximizes work output regardless of actuation stretch. Consequently, they share a low actuation speed limiting their power output. VII. CONCLUSION This paper has investigated the ability of antagonistic DEAs to provide simpler and higher energy dense bistable actuators than ip-op devices. Two bistable antagonistic DEAs (linear and rotating) were designed using an experimentally validated BergstromBoyce analytical model. The following conclusions are drawn. 1) The low dead volume of antagonistic bistable DEAs leads to 10 higher energy and 4 higher power densities than ip-op devices, providing a reduction in active lm layers requirement and actuator size. The reduction in active lm layers further translates into potential reliability improvements. 2) By tuning the actuation stretch of the lm, a tradeoff between DEA energy and power density is present. In fact, a rotating conguration tuned for low actuation stretch has slightly inferior energy density but 10 higher power density than a similar conguration tuned for high actuation stretch. This tradeoff is valid for any DEAs using viscoelastic polymers. 3) The work suggests that antagonistic bistable actuators are promising for binary robotics and mechatronics because of their improved energy density. However, more technological development is still needed for the manufacturing of multilayered actuators to bring their energy outputs to practical levels.

Future work will explore the development of multistable actuators for use in practical binary devices. Also, the impact of antagonistic bistable DEAs design parameters on reliability and long-term aging of the polymer lm remains unknown and must be addressed. REFERENCES
[1] V. Anderson and R. Horn, Tensor arm manipulator design, Proc. ASME, vol. 67, pp. 112, 1967. [2] M. Lichter, V. Sujan, and S. Dubowsky, Computational issues in the planning and kinematics of binary robots, in Proc. 2002 IEEE Int. Conf. Robot. Autom., pp. 341346. [3] L. Petit, C. Prelle, E. Dor e, F. Lamarque, and M. Bigerelle, A fourdiscrete-position electromagnetic actuator: Modeling and experimentation, IEEE/ASME Trans. Mechatronics, vol. 15, no. 1, pp. 88969, Feb. 2010. [4] M. Hafez, M.D. Lichter, and S. Dubowsky, Optimized binary modular recongurable robotic devices, IEEE/ASME Trans. Mechatronics, vol. 8, no. 1, pp. 1825, Mar. 2003. [5] Y. Bar-Cohen, Electroactive Polymer (EAP) Actuators as Articial Muscles: Reality, Potential, and Challenges. Bellingham, WA: SPIE, 2004. [6] G. Kofod, Dielectric elastomer actuators, Ph.D. dissertation, Dept. Chem., Techn. Univ. Denmark, Denmark, 2001. [7] F. Carpi, G. Frediani, and D. De Rossi, Hydrostatically coupled dielectric elastomer actuators, IEEE/ASME Trans. Mechatronics, vol. 15, no. 2, pp. 308315, Apr. 2010. [8] M. Wissler and E. Mazza, Electromechanical coupling in dielectric elastomer actuators, Sens. Actuators A, Phys., vol. 138, no. 2, pp. 38493, 2007. [9] P. Sommer-Larsen and A. Larsen, Materials for dielectric elastomer actuators, in Proc. SPIEInt. Soc. Opt. Eng., San Diego, CA, 2004, pp. 68 77. [10] R. Pelrine, R. Kornbluh, and G. Kofod, High-strain actuator materials based on dielectric elastomers, Adv. Mater., vol. 12, no. 16, pp. 1223 1225, 2000. [11] R. Kornbluh, Ultrahigh strain response of eld-actuated elastomeric polymers, Proc. SPIE, vol. 3987, pp. 5163, 2000. [12] J. Plante and S. Dubowsky, Large-scale failure modes of dielectric elastomer actuators, Int. J. Solids Struct., vol. 43, no. 25, pp. 77277751, 2006. [13] J. Plante, Dielectric elastomer actuators for binary robotics and mechatronics, Ph.D. dissertation, Dept. Mech. Eng., Massachusetts Inst. Technol., Cambridge, MA, 2006. [14] L. Devita, An MRI compatible manipulator for prostate cancer detection and treatment, M.S. dissertation, Dept. Mech. Eng., Massachusetts Inst. Technol., Cambridge, MA, 2007. [15] A. Wingert, M. D. Lichter, and S. Dubowsky, On the design of large degree-of-freedom digital mechatronic devices based on bistable dielectric elastomer actuators, IEEE/ASME Trans. Mechatronics, vol. 11, no. 4, pp. 448456, Aug. 2006. [16] J. Plante and S. Dubowsky, On the performance mechanisms of dielectric elastomer actuators, Sens. Actuators A, Phys., vol. 137, no. 1, pp. 96109, 2007. [17] R. Mutlu and G. Alici, A multistable linear actuation mechanism based on articial muscles, J. Mech. Design, vol. 132, no. 11, pp. 111001-1 111001-8, Nov. 2010.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
CHOUINARD AND PLANTE: BISTABLE ANTAGONISTIC DEAs FOR BINARY ROBOTICS AND MECHATRONICS 9

[18] M. Wissler, Modeling dielectric elastomer actuators, Ph.D. dissertation, Swiss Federal Instit. Technol., Zurich, Switzerland, 2007. [19] J. Bergstrom and M. Boyce, Constitutive modeling of the large strain time-dependent behavior of elastomers, J. Mech. Phys. Solids, vol. 46, no. 5, pp. 931954, May 1998. [20] M. Wissler and E. Mazza, Modeling of a pre-strained circular actuator made of dielectric elastomers, Sens. Actuators A, Phys., vol. 120, no. 1, pp. 184192, Apr. 2005. [21] R. Palakodeti and M. Kessler, Inuence of frequency and prestrain on the mechanical efciency of dielectric electroactive polymer actuators, Mater. Lett., vol. 60, no. 29, pp. 34373440, 2006. [22] J. Lucking Bigu e, P. Chouinard, S. Proulx, G. Miron, and J. Plante, Preliminary assessment of manufacturing impacts on dielectric elastomer actuator reliability, presented at the Cansmart, p. 12, 2009. [23] P. Lochmatter and G. Kovacs, Design and characterization of an active hinge segment based on soft dielectric EAPs, Sens. Actuators A, Phys., vol. 141, no. 2, pp. 577587, Feb. 2008. [24] G. Kofod, P. Sommer-Larsen, R. Kornbluh, and R. Pelrine, Actuation response of polyacrylate dielectric elastomers, J. Intell. Mater. Syst. Struct., vol. 14, no. 12, pp. 787793, Dec. 2003.

Jean-S ebastien Plante received the B.Sc.A. and M.Sc.A. degrees in mechanical engineering from the Universit e de Sherbrooke, Sherbrooke, QC, Canada, in 1998 and 2001, respectively, and the Ph.D. degree from the Massachusetts Institute of Technology (MIT), Cambridge, in 2006. Following postdoctoral studies in MITs Field and Space Robotics Laboratory, he joined the Department of Mechanical Engineering, Universit e de Sherbrooke, as an Associate Professor in 2007, where he founded the Conception dActionneur et de Moteur de lUniversit e de Sherbrooke laboratory, a laboratory dedicated to the development of new actuator and engine technologies for robotics and transportation. He has an eclectic approach and studies actuation technologies ranging from shock wave hydrogen engines to polymer actuators. His current research interests in mechatronics include high-energy and power-density actuators based on dielectric elastomers and pneumatic air muscles. He develops applications of these actuators to MRI-compatible medical devices.

Patrick Chouinard received the B.Sc.A. and M.Sc.A. degrees in mechanical engineering from the Universit e de Sherbrooke, Sherbrooke, QC, Canada, in 2008 and 2010, respectively. He is currently a member of the Conception dActionneur et de Moteur de lUniversit e de Sherbrooke laboratory team. His current research interests include mechanical design of high-performance actuators and system dynamics modeling.