A Review of the Applications and Challenges of Dielectric Elastomer Actuators in Soft Robotics

Abstract

As an electrically driven artificial muscle, dielectric elastomer actuators (DEAs) are notable for their large deformation, fast response speed, and high energy density, showing significant potential in soft robots. The paper discusses the working principles of DEAs, focusing on their reversible deformation under electric fields and performance optimization through material and structural innovations. Key applications include soft grippers, locomotion robots (e.g., multilegged, crawling, swimming, and jumping/flying), humanoid robots, and wearable devices. The challenges associated with DEAs are also examined, including the actuation properties of DE material, material fatigue, viscoelastic effects, and environmental adaptability. Finally, modeling and control strategies to enhance DEA performance are introduced, with a perspective on future technological advancements in the field.

1. Introduction

With continuous technological advancements, automation and intelligent systems are playing increasingly critical roles in our daily lives and work. Among various automated systems, robotics is particularly prominent, with applications that have expanded from traditional manufacturing to sectors such as services, healthcare, exploration, and national defense. Although traditional robots have achieved significant success in industrial settings due to their precision and repeatability, they are typically constructed using rigid materials, limiting their adaptability and flexibility in complex environments.

The biodiversity in nature offers researchers an endless source of inspiration. The flexibility and adaptability of living organisms have inspired the design of soft robots. These robots are constructed from soft materials and are capable of mimicking the movements and behaviors of natural organisms, demonstrating exceptional compliance and sensitivity to environmental changes. These characteristics of soft robots provide them with unique advantages in fields such as search, rescue, medical surgery, and flexible assembly.

There is a wide variety of artificial muscles, including but not limited to electroactive polymers (EAPs) [1,2,3], hydrogels [4,5,6,7], magnetic materials [8,9,10], shape memory alloys (SMAs) [11,12,13,14], shape memory polymers (SMPs) [15,16,17], liquid crystal elastomers (LCEs) [18,19,20], and ionic polymer-metal composites (IPMCs) [21,22,23]. These materials can deform when exposed to external triggers such as electric or magnetic fields, temperature transforms, and light, thereby enabling actuation. In addition, pneumatic actuation [24,25] is another commonly used method in soft robotics. Table 1 lists the optimal performance of various flexible actuation methods. Although these optimal performances were obtained from different actuators, with some even being theoretical estimates, they still reveal the potential upper limits of various flexible actuation technologies.

Notably, dielectric elastomers (DEs) stand out as a highly efficient type of EAP, offering advantages such as high energy density [26,27,28], fast response time [29,30], large deformations [31,32], high flexibility [33,34], and low noise and power consumption [35,36]. These attributes have led to extensive research and applications of DEs in soft robotics, earning them the reputation of being the next-generation “artificial muscles”.

This review delves into the latest research advancements in DEAs and their usage in soft robotics. The review commences with an introduction to the working principles of DEAs and the key material properties that underpin their functionality. Following this, the characteristics of various DEA design configurations and their diverse applications in soft robotics are discussed in detail. Additionally, the obstacles faced by DEAs in practical applications are investigated, along with an exploration of potential future developments in the field. In this review, we aim to offer readers a thorough insight into the significance and potential applications of DEAs in soft robotics, while inspiring further research and innovative ideas.

Table 1. Actuation technique performance overview.

Actuators	Strain (%)	Stress (MPa)	Work Density (KJ/m3)	Efficiency (%)	Relative Speed	References
Natural muscle (peaks in nature)	100	0.8	40	40	Fast	[37]
Natural muscle (human skeletal)	40	0.35	70	40	Medium	[37,38]
SMAs	10	200	10,000	10	Slow	[38,39]
SMPs	400	4	2000	10	Slow	[40]
Pneumatic actuators	99.7	1.16	100,000	49	Medium	[39,41,42]
DEAs	>500	7.7	3400	90	Fast	[39,43,44]
Ionic gels	40	0.3	60	30	Slow	[37]
IPMCs	40	3	5.5	1.5	Slow	[39]
Piezoelectric ceramic	0.2	110	110	90	Fast	[37]

Table 1. Actuation technique performance overview.

Actuators	Strain (%)	Stress (MPa)	Work Density (KJ/m3)	Efficiency (%)	Relative Speed	References
Natural muscle (peaks in nature)	100	0.8	40	40	Fast	[37]
Natural muscle (human skeletal)	40	0.35	70	40	Medium	[37,38]
SMAs	10	200	10,000	10	Slow	[38,39]
SMPs	400	4	2000	10	Slow	[40]
Pneumatic actuators	99.7	1.16	100,000	49	Medium	[39,41,42]
DEAs	>500	7.7	3400	90	Fast	[39,43,44]
Ionic gels	40	0.3	60	30	Slow	[37]
IPMCs	40	3	5.5	1.5	Slow	[39]
Piezoelectric ceramic	0.2	110	110	90	Fast	[37]

2. Dielectric Elastomer Actuators

As one of the key technologies for flexible actuation, DEAs provide a novel actuation mechanism with inherent compliance and bio-inspired features for soft robots. The following sections will analyze the working principles of DEAs, explore the properties of dielectric elastomer materials and compliant electrodes, and investigate how these characteristics combine to create actuators with multifunctionality and high efficiency.

2.1. Working Principle

The operation of DEAs depends on the reversible deformation of DE materials under the influence of a high electric field. Typically, DEAs are composed of a soft dielectric elastomer membrane sandwiched between compliant conductive electrodes. Upon applying a voltage to the electrodes, an electric field is generated within the membrane, producing electrostatic pressure (Maxwell stress) between the top and bottom electrodes. This pressure compresses the membrane in the thickness direction. Since dielectric elastomers are nearly incompressible, this compression results in lateral and longitudinal expansion, leading to significant mechanical deformation, as illustrated in Figure 1a. Upon removal of the voltage, the DEAs return to their original shapes owing to their inherent elasticity, allowing for repeated actuation cycles. Due to the high flexibility and low density of dielectric elastomer materials, DEAs offer fast response and low energy consumption during actuation. This allows DEAs to be designed as various types of actuators and assembled into a variety of soft robots [45,46,47,48]. Conversely, as illustrated in Figure 1b, DEAs can also serve as energy harvesters by converting mechanical energy into electrical energy [49,50]. Furthermore, their capacitor-like structure enables deformation-induced changes in capacitance, as depicted in Figure 1c, which allows DEAs to be used as flexible sensors [51,52]. Since DEAs are self-driving and self-sensing, the closed-loop control method can be applied to DEAs without external sensors [53].

During the actuation process, a DEA operates similarly to a flexible capacitor, capable of transforming electrical energy into mechanical motion. Its capacitance C is determined by the following formula:

$$C={\displaystyle \frac{{\epsilon }_0\epsilon A}{z}}$$

(1)

In this formula, $\epsilon$ represents the relative permittivity of the dielectric elastomer material. ${\epsilon }_0$ is the permittivity of the vacuum. A is the area covered by the compliant electrode and z is the thickness of the DE film. The electrical energy u stored on the charged DEA can be expressed as

$$u={\displaystyle \frac{0.5Q^2}{C}}={\displaystyle \frac{0.5z{Q}^2}{{\epsilon }_0\epsilon A}}$$

(2)

where Q denotes the amount of charge stored on the DE film. Since DE materials are inherently incompressible, we can substitute the volume constant into the equation and obtain the Maxwell stress p acting on the two electrodes as

$$p={\epsilon }_0\epsilon E^2={\epsilon }_0\epsilon {({\displaystyle \frac{V}{z}})}^2$$

(3)

where E is the electric field intensity and V is the applied voltage. In the Maxwell stress equation, it is assumed that the DE material thickness is uniform and that the charge is also uniformly distributed. Based on the free boundary conditions and the assumption of linear elasticity for small strains (<10%), the thickness strain S_z is given by Reference [54] as follows:

$$S_z=-{\displaystyle \frac{p}{Y}}=-{\displaystyle \frac{{\epsilon }_0\epsilon E^2}{Y}}=-{\displaystyle \frac{{\epsilon }_0\epsilon }{Y}}{({\displaystyle \frac{V}{z}})}^2$$

(4)

where Y is the elastic modulus of the material. In this way, the actual size after deformation can be expressed by the initial size through its constant volume property, that is, the thickness is expressed as $z=z_0(1+S_z)$, so that the deformation in the plane can be obtained as S_x and S_y as

$$(1+S_z)(1+S_y)(1+S_x)=1$$

(5)

For isotropic dielectric elastic materials, let $S_a=S_x=S_y$. Formula (5) can be written as

$$S_a={(1+S_z)}^{-0.5}-1$$

(6)

From Formulas (4)–(6), it can be seen that the deformation of a DEA is related to elastic modulus Y and the relative dielectric constant Ɛ, but these two parameters are related to the material properties, temperature, actuation frequency, stress state, and material fatigue. In addition, since most DE materials have nonlinear and viscoelastic properties that are inconsistent with the linear elastic assumption of this model, a more accurate model is needed to predict the dynamic behavior of DEA. The operational capabilities of DEAs primarily depend on the properties of the DE and the compliant electrodes.

2.2. Properities of DE Materials

The functionality of a DEA is largely governed by its core component—the dielectric elastomer material. For optimal actuation performance, the DE material must possess a low elastic modulus, minimal viscoelasticity, a high permittivity, and robust dielectric strength to endure high electric fields without breaking down.

In previous studies, DEAs have predominantly utilized acrylic and silicone elastomers as their primary dielectric materials. Specifically, acrylic elastomers, such as VHB 4910 and VHB 4905 from 3M Company (St. Paul, MN, USA), have been favored in experimental settings due to their widespread industrial application, cost-effectiveness, and their capacity to achieve significant electro-induced strain, with linear strains surpassing 380% [37] under substantial pre-strain conditions and area strains over 2200% [55]. However, due to the viscoelasticity and nonlinearity of the acrylic elastomers, the response speed of a DEA is slow [56]. Compared with acrylic elastomers, the low viscoelasticity of silicone elastomers makes silicone elastomers respond faster in dynamic applications, with lower mechanical losses, and can be used at higher operating frequencies [57].

Table 2. Property overview of DE materials.

Type		Pre-Strain (x,y) (%)	Maximum Area Strain (%)	Young’s Modulus (MPa)	Electric Field (V/μm)	Dielectric Permittivity	References
Acrylic polymer	3M VHB 4910	(300,300)	158	3	412	4.8@1 kHz	[58]
	3M VHB 4910	(540, 75)	215	-	239	4.8@1 kHz	[58]
	VHB-poly (TMPTMA)	(400, 400)	300	4	420	-	[59]
	VHB-IPN-P	(250, 250)	185	0.48	66	-	[60]
	PHDE	-	189	1.3	120	5.35@1 kHz	[26]
	PIL/VHB 4905	(200, 200)	133	0.21	17	16.4@1 kHz	[61]
	PUA-PEGDA-15	-	71.4	0.323@10%	24.2	9.4@1 kHz	[62]
Silicone polymer	Nusil CF19-2186	(45, 45)	64	1	350	2.8@1 kHz	[58]
	Dow Corning HS3	(68, 68)	93	0.1	93	2.8@1 kHz	[58]
	Bottlebrush polymer based on PDMS	-	>300	-	<10	-	[63]
	exo-E62′	-	12	0.019@10%	15	3.02@10 kHz	[64]
	SR5 (5% 81-R/ silicone elastomer)	(40, 40)	10	0.35	32	3.25	[65]
	C2	-	20	0.154@10%	10.8	10.1@10 kHz	[66]
	E-CL2-3	-	14	0.628@10%	24.2	18.4@10 kHz	[67]
	90 phr BmimSbF6/ silicone elastomer	-	-	0.15	7.5	[email protected] Hz	[68]

shows the current research progress on major dielectric elastic materials.

Nowadays, DE materials have gradually become a research hotspot owing to their prospective utility in soft robotics and other domains. Through continuous material innovation and design optimization, DE materials are poised to play a pivotal role in the development of future intelligent systems.

2.3. Compliant Electrodes

Compliant electrodes play a vital role in DEAs. They must possess a series of key properties to meet the needs of applications. First, they need to be highly stretchable so that they can adapt to the large deformation of dielectric elastomers under the influence of electric fields. At the same time, they should maintain high conductivity to ensure the effective transfer of electrical energy and fast response. Desirable compatibility and strong adhesion are also essential to ensure a close bond between the electrode and the DE material to avoid peeling or damage during repeated use. Stability and durability are two other important properties of stretchable electrodes which need to maintain good performance during long-term applications and resist the influence of environmental factors such as humidity, temperature changes, and ultraviolet radiation. Some advanced stretchable electrodes also have self-healing capabilities, which allows them to restore their conductivity and mechanical properties after damage, thereby extending the service life of the device [69].

At present, common types of stretchable electrodes include carbon-based electrodes (such as carbon black, graphite, and carbon nanotubes) [31,35,70,71,72,73,74] and silver nanowires [75,76]. These materials are integrated onto soft substrates through different processing techniques such as coating, spraying, chemical vapor deposition, or 3D printing. Among them, carbon grease is the most widely applied flexible electrode due to its low cost and excellent compliance. However, a major drawback is that the oil in the carbon grease tends to dry out over time. The advantage of graphite electrodes is their good conductivity, but their flexibility is poor. Under large deformation, they will have poor contact with the elastomer and affect the conductivity [77,78]. Silver nanowires have high conductivity and flexibility, and their good transparency makes them very useful for some optical applications (such as transparent displays or flexible optoelectronic devices), but they are expensive and prone to oxidation problems when used for a long time or in harsh environments [40]. Carbon nanotube (CNT) films exhibit a unique self-clearing capability after breakdown damage, which represents a significant advancement in extending the lifespan of DEAs [31,34]. However, this self-clearing process inevitably burns off portions of the CNT electrodes, resulting in a reduction in actuation strain. Moreover, during fabrication, CNTs tend to agglomerate, making it challenging to achieve uniform dispersion, which can compromise the stability of the electrodes. Additionally, due to the limited adhesion of CNTs, high-frequency actuation can lead to detachment and poor contact, further affecting performance.

2.4. Typical Configurations of DEAs

DEAs have simple structures, comprising two flexible electrodes with a dielectric elastomer layer in between. This simplicity enables the creation of DEAs tailored to specific application needs through various modifications. Most DEAs are designed based on the area expansion of dielectric elastomer films. Nevertheless, there are also many multi-layer stacked or folded DEAs, which are actuated by the thickness reduction in the dielectric elastomer after a high voltage is imposed. Depending on the different actuation forms of a DEA, scholars have designed many DEA configurations, as illustrated in Figure 2.

As shown in Figure 2a, a sheet bender involves attaching sheet-like DEAs to a soft, non-stretchable film. When a high voltage is applied to the DEA, it expands in the area, and due to the constraint of the non-stretchable film, the entire mechanism bends unidirectionally in the opposite direction [79]. Similarly, by attaching DEAs to both sides of the film, bidirectional bending can be achieved [80]. Figure 2b illustrates the structure of a rolled actuator, which is primarily formed by rolling the DEA into a cylindrical shape [81,82], sometimes wrapping it around a spring [83,84,85]. When a high voltage is applied, the expansion of the DEA’s area causes the actuator to elongate in length. In practical applications, stretchable electrodes can be arranged on one side and then rolled, enabling the actuator to achieve unilateral elongation, thereby generating a bending effect [86]. As shown in Figure 2c, a multi-layer actuator is constructed by stacking multiple thin DEA layers [72,81,87]. This structure is equivalent to a parallel arrangement of multiple actuators, aiming to achieve a greater driving force under the same operating voltage [88]. A tubular DEA is shown in Figure 2d, which differs from rolled DEAs in that its ends are typically fixed, with a hollow center to allow the passage of various fluids [32,89]. When voltage is applied, the actuator’s area expands, causing external fluid to be drawn in. Upon removal of the voltage, the actuator returns to its original shape, expelling the fluid accordingly. A balloon-shaped actuator, as shown in Figure 2e, is typically connected to an air valve to maintain internal pressure and guide the deformation of the DEA [90,91,92]. When a high voltage is applied, the expansion of the DEA’s area results in an increase in the balloon’s volume. Figure 2f illustrates a regionally segmented actuator [27], wherein a pre-stretched DE film is partitioned into discrete sections, each independently coated with flexible electrodes. When a specific region is electrically activated, it undergoes localized expansion, which correspondingly results in the contraction of the opposing region.

3. Utilization of DEAs in Soft Robotics

Over the past few years, soft robots have gained increasing attention from scholars owing to their unique flexibility and compliance. The simple structure of DEAs allows soft robots to more effectively emulate natural motions, opening up new possibilities for their design and applications. The following sections focus on the use of DEAs in soft robotics, including soft grippers, multilegged robots, crawling robots, swimming robots, jumping/flying robots, humanoid robots, and wearable devices.

3.1. Soft Grippers

Grasping objects of different shapes and types has consistently been a challenging issue in the field of robotics. Traditional rigid grippers excel at providing high output forces and handling heavy loads but often struggle with irregularly shaped, soft, or fragile objects, where they risk causing damage. Due to their inherent flexibility and compliance, DEAs offer an effective solution to this type of problem. Researchers have developed various soft gripping devices based on DEAs, as shown in Figure 3.

Earlier soft grippers were often based on dielectric elastomer minimum energy structures (DEMES) [93]. These structures are composed of a pre-stretched dielectric elastomer film attached to a flexible frame, typically made of polyethylene terephthalate (PET). While the PET frame is flexible, it cannot be stretched, causing the structure to bend passively when released. By coating electrodes on both surfaces of the dielectric elastomer membrane and applying a high voltage to the active area, the stress in the structure is released. This enables the actuator’s planar deformation to be converted into out-of-plane deformation in the gripper, while simultaneously minimizing the actuator’s out-of-plane displacement. Kofod et al. [93] introduced a tulip-shaped soft gripper based on DEMES, as illustrated in Figure 3a. When voltage is imposed, the actuator causes the three claws to open and envelop the object. Once the voltage is removed, the gripper contracts to securely clamp the object. This design requires continuous energy consumption to maintain the gripper in an open state.

Inspired by the Venus flytrap, Wang et al. [94] designed a bistable soft gripper founded on DEMES, as shown in Figure 3b. The gripper features two PET frames with pre-stretched DE films attached on both sides. Upon applying voltage, the gripper transitions between two stable states in just 0.17 s, enabling the rapid grasping and releasing of objects. Notably, each grasping action consumes only 0.1386 J of energy, and no continuous power supply is required to maintain the grasping state. Additionally, Lau et al. [86] developed a flexible gripper using pre-stretched VHB and a DEMES-based design. As shown in Figure 3c. By applying a high voltage to the actuators, the gripper can transition from an initial curved (passive) state to a straight (active) state. It is capable of achieving voltage-controlled bending up to 90°, with a high bending stiffness, allowing it to lift a load 8–9 times its own weight when fully closed. Additionally, the gripper can precisely pick up delicate objects like raw egg yolks when fully opened. Under an activation voltage of 6 kV, the gripper can generate a maximum blocking force of 143 mN, approximately six times the weight of a single finger.

In addition to DEMES, other DEA configurations have also been universally applicated in soft grippers. Dou et al. [95] designed a soft robotic gripper based on balloon-shaped DEAs, with the dielectric layer composed of poly (laurolactam acrylate) (PLA). As illustrated in Figure 3d. The large side groups in the PLA network mutually repel each other, extending the network chains that bear the load, thereby imparting significant strain-hardening behavior to the material. This behavior helps to prevent electromechanical instability (EMI) during electrical activation, achieving up to 218% voltage-induced deformation. The material also exhibits an energy density of up to 244.2 kJ/m³. The gripper is capable of grasping, lifting, and rotating fragile objects of various complex shapes and different weights, such as ping-pong balls, shuttlecocks, strawberries, toy ducks, and grapes. Despite its total weight being approximately 200 mg, the gripper can generate a gripping force exceeding 1 N (about 500 times its own weight), with a stable gripping force that is unaffected by the surface roughness of the object. Moreover, the gripper exhibits rapid response characteristics, enabling it to capture a freely falling ping-pong ball, with an instantaneous speed of approximately 3000 mm/s at the moment of capture. Aksoy et al. [96] designed a gripper capable of directional bending, as illustrated in Figure 3e. The gripper is made up of a multi-layer structure, including a DEA, two layers of shape memory polymer (SMP) fibers, and a series of stretchable heaters. The DEA serves as the active layer of the gripper, while the SMP fibers create localized soft regions when heated, allowing the DEA to bend along these soft axes upon voltage application. The gripper can achieve a tip deflection angle of over 300° and exhibits a blocking force of more than 27 mN at 5 kV. Shintake et al. [79] introduced a soft gripper underpinned by a multifunctional polymer actuator, as shown in Figure 3f, which uses a new electrode arrangement and has an intrinsic electrosorption capability. At zero voltage, the gripper remains in a curled position. Upon applying voltage, the fingers close around the object, generating an electrosorption force. Compared with its electrosorption force (which can generate 3.5 N shear force), its mechanical gripping force is very small (1 mN), which allows the gripper to grasp and release objects of various shapes and weights, including fragile, highly deformable water balls, raw eggs, paper, metal cans, etc.

3.2. Multilegged Robots

In the early exploration of DEA applications in soft robotics, Eckerle et al. [97] introduced the first DEA-actuated six-legged walking robot, FLEX, in 2001. FLEX weighs 670 g and uses bowtie-shaped actuators made of acrylic elastomer, with each leg providing two degrees of freedom (up/down and forward/backward). Although FLEX was able to walk at a speed of several millimeters per second, its movement was considered too slow. Later, in 2002, the same team introduced FLEX 2 [98], an improved hexapod walking robot that retained the basic kinematic design of the original FLEX but featured enhancements for better performance. FLEX 2 uses a more powerful rolled DEA. It achieved a significant improvement in walking speed, increasing from a negligible few millimeters per second to a remarkable 3.5 cm/s. Additionally, its lifespan and shelf life were greatly enhanced. A key feature of both FLEX and FLEX 2 is the integration of battery power supplies, voltage conversion systems, and microprocessor controllers, eliminating the need for external signal generators and high-voltage amplifiers. These advancements make the FLEX series a landmark achievement in the development of DEA-actuated biomimetic robots.

Pei et al. [83,84] also designed two other DEA-actuated multilegged robots, Skitter and MERbot. Skitter is a compact multi-legged walking robot. Its six legs are made of rolled single-DOF DEA. Skitter’s maximum speed is about 7 cm/s. Unlike Skitter, MERbot’s six legs are made of 2-DOF spring-roll DEA, and the maximum bending angle of each leg is 90°, which enables the robot to walk in a tripod gait. As shown in Figure 4a, MERbot has a size of 18 cm × 18 cm × 10 cm and reaches its maximum speed (13.6 cm/s) at 7 Hz and 5.5 kV, which is about two-thirds of its body length.

In 2014, Nguyen et al. [99] designed a compact bionic quadruped robot actuated by multi-layer stacked DEAs, as illustrated in Figure 4b. Each leg of the robot is fitted with two DEAs, enabling movement during both the swing and stance phases. With overall dimensions of 191 mm × 100 mm × 115 mm and a weight of 450 g, the robot is capable of performing both walking and running, as demonstrated by experimental results. Later, the same research group arranged the multi-DOF conical DEAs in an antagonistic manner and developed a hexapod walking robot [100]. As depicted in Figure 4c, the robot is made up of six 3-DOF DEAs, which can provide two-dimensional translation and one-dimensional rotational motion through their antagonistic action. The robot has a size of 150 mm × 106 mm × 48 mm and weighs approximately 35 g in total. It can use an alternating tripod gait to walk forward and backward in a straight line. The walking speed is 30 mm/s (0.2 BL/s) at an actuating frequency of 2 Hz, and it can carry a maximum load of 1.4 times its own weight to complete walking. In 2018, the research group further improved the robot and developed another hexapod robot using three 5-DOF DEAs [101] as shown in Figure 4d. Each actuator can provide three translational movements and two rotational movements, which enables the robot to not only walk forward and backward in a straight line but also walk sideways and turn. The robot’s structure has become more compact (size is 150 mm × 54 mm × 55 mm, weight is 20 g), and its motion performance has also been improved. The average walking speed at 7 Hz can reach 5.2 cm/s, about 0.3 BL/s.

3.3. Crawling Robots

Within the realm of soft robotics, imitating the movement of natural organisms has always been an important research direction. For DEA-driven soft crawling robots, learning from the movement patterns of reptiles and insects such as inchworms and caterpillars is of great significance for designing soft robots that can move flexibly in intricate environments. Gu et al. [102] developed a soft wall climbing robot as illustrated in Figure 5a. The robot uses the rapid periodic deformation of DEA and the controllable adhesion of its electroadhesive feet to achieve effective crawling on the surfaces of various materials. It is particularly worth mentioning that the robot achieved a crawling speed of 63.43 mm/s (0.75 BL/s) on the wooden surface, which is excellent among similar soft robots. In addition, it can carry a payload equivalent to a certain proportion of its own weight. For example, when carrying a 10 g load, it can still crawl steadily at a speed of 1 mm/s on a wooden surface. Tang et al. [103] imitated the movement pattern of earthworms and designed a sub-centimeter-scale modular pipe crawling robot as illustrated in Figure 5b. The robot is composed of multiple anchoring and elongation mechanisms. The components are connected by small magnets. The number of units can be easily adjusted to adapt to different pipe shapes. The robot can achieve rapid horizontal and vertical movement in a 9.8 mm-diameter pipe (horizontal: 1.19 BL/s; vertical: 1.08 BL/s). In addition, the robot can also move in different materials (glass, metal, or carbon fiber) filled with air or oil, and can adapt to pipes with varying diameters.

Du et al. [104] designed a crawling robot based on a spring-rolled dielectric elastomer (SRDE), inspired by the inchworm’s movement pattern, as illustrated in Figure 5c. The crawling robot is primarily made up of a single-degree-of-freedom spring-roll actuator and two bristles. Thanks to the flexibility of the front and rear bristles, the robot demonstrates good adaptability in pipes with different diameters. In pipes with an inner diameter smaller than the bristle’s outer diameter, it relies on anisotropic friction to achieve crawling, with horizontal and vertical crawling speeds of 1.88 BL/s and 0.88 L/s, respectively. In pipes with an inner diameter larger than the bristle outer diameter, crawling is facilitated by small bending moments generated by changes in body length, achieving a speed of up to 2.78 BL/s. Ji et al. [105] introduced a novel crawling robot named DEAnsect, as illustrated in Figure 5d. DEAnsect is a soft robot actuated by low-voltage stacked dielectric elastomer actuators (LVSDEAs), weighing only 780 mg and measuring 40 mm in length. Each leg of DEAnsect is independently actuated by an LVSDEA, utilizing directional friction forces for locomotion. Operating at a driving voltage of 450 V and a frequency exceeding 600 Hz, DEAnsect achieves a maximum speed of 30 mm/s. Its soft body allows it to withstand extreme deformations, for instance, allowing it to continue to function after being struck by a fly swatter, demonstrating exceptional robustness. Moreover, DEAnsect integrates sensors, controllers, and batteries, enabling the miniaturization of its control system while achieving autonomous intelligent navigation. Wang et al. [106] developed a new type of DEA soft robot, as illustrated in Figure 5e. The robot demonstrated the ability to instantly change the direction of movement during rapid movement through its unique chiral lattice foot design. The robot uses a DEA as its deformable body and uses the adjustment of voltage frequency to control its multimodal motion. Experiments have demonstrated that the robot can move backward at speeds of up to 124 mm/s, forward at 112 mm/s, and perform circular motion with an angular velocity of approximately 0.37 radians/s. In addition, by combining structural design and the shape memory effect of smart materials, the robot is able to achieve complex S-shaped trajectories or reduce its height to pass through narrow spaces. Zhu et al. [107]. introduced a 3D-printed insect-scale soft robot, as shown in Figure 5f. The robot uses high-frequency voltage to drive DEA and demonstrates the ability to move at ultra-fast speeds, with a speed of nearly 4 BL/s. This robot is not only small in size and light in weight, but it also has high robustness and good environmental adaptability. Manufactured through an ingeniously designed five-inlet multi-material coaxial 3D printing technology, the DEA can work stably at high frequencies and have a long life of more than one million cycles. In addition, the robot also has the ability to move in complex environments, including the ability to move quickly on varied surfaces, crawl in narrow pipes, and work collaboratively in a robot group. Ma et al. [108] developed a new type of soft cable crawling robot, as shown in Figure 5g, which is designed for cable inspection in narrow spaces. This centimeter-sized robot achieves fast and efficient crawling motion through three multilayer DEAs. Its crawling speed reaches 0.72 BL/s, making it one of the fastest cable crawling robots currently available. The robot can not only crawl on cables of various materials and diameters, but it can also carry objects weighing up to 3.69 times its own weight and has the ability to crawl on the water–air interface.

The working environment of DEA-driven crawling robots is no longer limited to crawling in a single working environment. Soft crawling robots that can adapt to multiple terrains and multiple working environments have gradually been developed. Wang et al. [109] developed an ultra-thin soft robot (TS-Robot) as shown in Figure 5h. It is only 1.7 mm thick and can adapt to multi-modal motion in narrow spaces. These robots use an innovative double-activated layer sandwich structure and an adjustable Poisson’s ratio tensioning mechanism to achieve the conversion of linear and wave gaits. Experiments show that TS-Robots can crawl, climb, turn, and swim in solid and liquid domains, demonstrating their adaptability and motion capabilities on a variety of substrates. In addition, these robots can also work in conjunction with other types of robots. such as drones, to expand their motion modes in complex environments, providing new solutions for the application of robots in the area of slit exploration. Cheng et al. [110] introduced a highly robust amphibious soft robot (AISR), as shown in Figure 5i. This robot uses a new type of water-stable and self-healing ionic conductor. Through precise molecular design, this ionic conductor material shows stability in both underwater and terrestrial environments and significantly improves the work life of the robot. The AISR can move quickly on land, with a speed of more than 4 BL/s, and can carry a load weighing up to 3.35 times its own weight. Underwater, the AISR can move with a velocity of 1.2 BL/s and can hold a load equivalent to seven times its own weight. In addition, the adaptability of AISR in mixed water and land environments, as well as its high stealth, gives it a wide range of potential applications in complex environments.

3.4. Swimming Robots

Due to the characteristics of dielectric elastomers, such as light weight, density close to water, and good transparency, a DEA has unique advantages in the application of bionic swimming robots. At present, scientists have developed a variety of DEA-driven swimming robots, as illustrated in Figure 6. Inspired by the movement of squids. Yang et al. [111] developed a DEA-driven robot as shown in Figure 6a. Furthermore, researchers utilize reinforcement learning to improve the control approach. The robot uses conical DEA to provide an actuation force and achieves untethered underwater movement. With the integrated jet propulsion, the robot can achieve compact movement similar to a motor drive system. With the help of reinforcement learning, the robot’s swimming speed was significantly improved to 21 mm/s (0.38 BL/s).

In 2018, Cheng et al. [112] developed a highly adaptable, untethered bionic jellyfish robot, as shown in Figure 6b. To improve autonomy, the researchers embedded a small remote control power supply on the top of the robot, enabling the robot to operate independently without an external power supply. In performance tests, the robot demonstrated excellent athletic ability, reaching a peak swimming speed of 1 cm/s. In 2020, a self-powered soft robot capable of autonomous operation in the extreme deep-sea environment of the Mariana Trench was introduced, reaching depths of up to 10,900 m [113], as shown in Figure 6c. This innovative design embedded electronic components within a silicone rubber matrix, providing a pressure-resistant structure that eliminated the need for traditional rigid pressure vessels. The robot’s decentralized electronic design reduced shear stress between components, enhancing its resilience to the extreme pressures of the deep sea. Utilizing specially designed DE materials for its flapping fins, the robot was able to achieve successful actuation even under the high-pressure conditions prevalent in the deep ocean. In field tests conducted in the South China Sea at a depth of 3224 m, the soft robot demonstrated impressive swimming capabilities, reaching speeds of 5.19 cm/s (0.45 BL/s). This achievement not only highlights the potential of soft robotics in extreme environments but also provides a new perspective for the design of future deep-sea exploration equipment, suggesting that soft, lightweight, and adaptable devices could play a crucial role in uncovering the mysteries of the deep sea.

In addition, Wang et al. [82] introduced a novel soft underwater robot, as illustrated in Figure 6d. When a high voltage is applied, the deformation of the DEAs causes the robot’s “legs” to rapidly form a semi-enclosed space, effectively increasing forward thrust. The specialized “adaptive foot” structure can intelligently adjust its deformation based on the direction of water pressure, altering the contact area with the water, thus enhancing thrust and reducing resistance. Experimental results show that the robot with adaptive feet exhibits a significant improvement in swimming speed, reaching a maximum velocity of 0.77 BL/s, and is capable of agile turning with a minimum turning radius of approximately 70 mm. Nagai et al. [85] introduced a DEA-based antagonistic actuator for propulsion in underwater robots as illustrated in Figure 6e. The actuator consists of a rigid skeleton, an elastic hinge, and two coiled DEAs arranged in an antagonistic manner. By alternately activating each DEA, bidirectional movement at the actuator’s tip is achieved. This motion is primarily inspired by the movement of fish fins in nature. In experiments, the swimming robot driven by this actuator demonstrated a forward speed of 0.9 mm/s under a voltage of 1000 V and a driving frequency of 10 Hz. Christianson et al. [114] introduced a translucent underwater robot as shown in Figure 6f. The design is inspired by the transparent leptocephali. The robot utilizes an internally liquid-filled chamber as one electrode, with the surrounding environmental liquid serving as the other electrode, simplifying the implementation of soft-bodied actuation devices. By mimicking the undulating swimming motion of eels, the robot achieves a maximum forward speed of 1.9 mm/s and a Froude efficiency of 52%. Its body has an average light transmittance of 94%, similar to that of transparent eel larvae, providing excellent camouflage and display capabilities.

3.5. Jumping/Flying Robots

Jumping is a widespread form of motion observed in living organisms. Completing the jumping action requires a rapid and effective stimulus response and sufficient actuation force. In earlier studies, Pei et al. [115] used DEAs to develop a jumping robot. This jumping robot has a simple structure and light weight. Three planar DEAs actuate the movement of the robot’s three legs individually. The jumping robot measures 5.5 cm × 5.5 cm × 1.5 cm. Upon applying voltage, the DEA produces out-of-plane motion, enabling the robot to leap, achieving leaps that are 1.33 times its own body height. Luo et al. [116] developed a jumping robot actuated by DEA as illustrated in Figure 7a. The researchers stacked multiple layers of dielectric elastomer films in parallel, creating a 20-layer DEA capable of delivering a peak output force of up to 30 N, significantly enhancing the driving performance of the DEA and enabling heavy-load actuation. The robot design consists of the following three parts: the DEA, the jumping mechanism, and the energy storage and release system. Through simulations and experimental validation, it was demonstrated that after 60 s of energy storage, the robot can achieve a jump, with a maximum jump height of 45 mm. Duduta et al. [117] introduced an electro-controlled latching compliant jumping mechanism based on a DEA, as illustrated in Figure 7b, aiming to provide a new solution for robotic mobility or self-recovery in unstructured environments. The design was inspired by the click beetle, an insect capable of generating jumping movements through rapid body folding without using leg strength. The researchers enhanced the traditional DEA framework by adding a stiffer strip, creating a double-layer composite beam. This composite beam can rapidly transition from a flat to a bent state under voltage control, thereby storing and quickly releasing mechanical energy to achieve the jump. Specifically, the DEA stretches when charged, reducing the bending of the composite beam. When the DEA rapidly discharges, the composite beam is quickly restored to its original bent state, propelling the central mass upward to achieve the jump. In experiments, this mechanism demonstrated the ability to achieve jumps of up to 5 cm.

DEA has also been used to develop flying robots based on biomimetic flapping wing mechanisms. Zhao et al. [118] introduced a flapping wing robot actuated by stacked DEAs, as shown in Figure 7c, and pointed out that although DE materials have sufficient power density to drive the flapping wing structure to take off, the current design has challenges in terms of overall power density. Lau et al. [119] incorporated rolled DEA into a lightweight carbon fiber reinforced polymer (CFRP) shell to maintain pre-stretching and improve the total actuation performance, as shown in Figure 7d. Although 30.9% of the theoretical working density was achieved, the wing could only produce 5–10° of travel when using the BJB-TC5005 ((BJB Enterprises, Inc., Irvine, CA, USA) membrane.

Chen et al. [69,120,121,122,123,124] and their research team developed a new type of flapping wing robot by using a novel form of rolled DEA, as shown in Figure 7e. They achieved high power density and high conversion efficiency for soft micro flapping wing aircraft, thus endowing the aircraft with insect-like flight capabilities and high maneuverability. These micro aircraft are not only capable of precise hovering flight but can also quickly recover after a collision and even perform complex flight maneuvers such as backflips. In addition, the team also innovatively introduced laser-assisted repair technology, which sufficiently enhanced the durability of the actuator and the reliability of the aircraft, as shown in Figure 7f. These research results not only demonstrate the potential of soft robots in the realm of flight but also provide important technical support and theoretical basis for the development of autonomous flying robots in complex environments in the future.

3.6. Humanoid Robots

The manufacture of humanoid robots has consistently posed a significant obstacle in robotics. Due to their muscle-like properties, including actuation strain, energy density, and response time, DE materials are considered among the most promising candidates for fabricating humanoid robots. Figure 8 illustrates several humanoid robots developed using DEA technology.

In an early study, Kovacs et al. [125] introduced an arm wrestling robot as illustrated in Figure 8a. Inspired by the antagonistic action of human muscles, more than 250 DEAs were organized into two groups to compete in an arm wrestling match with a human. Although the robot ultimately lost the match, it demonstrated the similarity between DEAs and human muscle movements. Later, Lu et al. [126] introduced an innovative fiber-constrained dielectric elastomer actuator that produced a linear unidirectional actuation strain of up to 142% by cleverly applying large pre-stretched rigid fibers. They also applied it to the development of a bionic artificial arm and successfully enabled a 70° rotation of the forearm in relation to the upper arm, as shown in Figure 8b. Furthermore, other researchers have utilized DEAs to mimic human arm movements, such as forearm lifting [30,72,127] and rotation [128]. These studies showcased the potential of DEAs in emulating human muscle functions.

Researchers have also tried to use DEAs to simulate human facial movements. Carpi et al. [129] used three antagonistic spring-roll DEAs to simulate the movement of eye muscles and achieved the horizontal movement of the humanoid robot eyeball, as illustrated in Figure 8c. And Luo et al. [130] used four DEMES to simulate eye muscles and achieved similar results. Wang et al. [131,132] used a sheet actuator to simulate the deformation of human jaw muscles as shown in Figure 8d. A feedforward controller was designed based on a viscoelastic nonlinear dynamic model, enabling the precise tracking of sinusoidal, triangular, and step trajectories. They also simulated the jaw movement in Dr. Martin Luther King’s famous speech “I Have a Dream”.

Figure 8. Humanoid robots based on DEAs: (a) arm wrestling robot [125]; (b) bionic robotic arm [126]; (c) humanoid robot eyeball [129]; (d) humanoid jaw muscles [131].

Figure 8. Humanoid robots based on DEAs: (a) arm wrestling robot [125]; (b) bionic robotic arm [126]; (c) humanoid robot eyeball [129]; (d) humanoid jaw muscles [131].

3.7. Wearable Devices

Tactile feedback is essential for robots to safely interact with humans or maneuver around obstacles in unpredictable environments. There has been much research on using DEA to develop wearable devices with tactile interaction capabilities. Ji et al. [133] introduced a tetherless haptic feedback device based on an 18-micron-thick DEA for tactile perception in virtual reality (VR) and augmented reality (AR), as illustrated in Figure 9a. The device directly adheres to the skin and can generate rich vibrational tactile feedback within a frequency range of 1 Hz to 500 Hz. The DEA deforms the elastomeric film by applying voltage between the electrodes, using electrostatic forces to stretch the skin and provide tactile feedback. Under a driving voltage of 450 V, the device achieves over 25% planar area strain, with a 14% strain at 10 Hz. As the frequency increases, the strain decreases due to the viscoelasticity of the silicone. The device’s RC time constant is approximately 2 ms, and with mechanical damping, the maximum operating frequency is 500 Hz. User tests showed that volunteers could recognize tactile signals of different frequencies and sequence patterns with a success rate ranging from 73% to 97%. Moreover, the tetherless version weighs only 1.3 g, making it suitable for prolonged wear. Lee et al. [134] introduced an array-based haptic display device utilizing a DEA as illustrated in Figure 9b. The device uses liquid coupling to transmit forces between the touch points and the actuators, ensuring comfort and safety during skin contact. By applying voltage to generate Maxwell stress, the actuators make contract in the thickness direction and expand radially, causing the touch layer to move up and down. The design of the device is based on Pascal’s principle, amplifying the actuator’s movement via liquid coupling without the need for complex mechanical transmission. Experimental results show that the device displaces approximately 240–120 μm within a frequency range of 3–10 Hz. It can generate forces exceeding 40 mN and simulate the tactile sensation of a fingertip. Additionally, the device incorporates a hydraulic amplification mechanism to enhance displacement. Performance improvements were achieved by altering the actuator’s substrate, increasing the electrode diameter, reducing the amount of liquid inside the device, and ensuring a smooth mold surface. Psychophysical experimental results indicate that the device is capable of accurately simulating fingertip sensations.

At the same time, there is also extensive research on the use of DEAs to develop flexible exoskeletons and wearable rehabilitation devices. Allen et al. [135] designed an active ankle–foot orthosis to address the issue of foot drop. The orthosis weighs just 1.30 kg and is composed of 81 DEAs, including 3 DEAs stacked in series and 27 DEAs arranged in parallel, as shown in Figure 9c. The device is used to provide sufficient dorsiflexion support to lift the foot without increasing volume, noise, and energy consumption. Experiments have proven that the device can provide the required 49% dorsiflexion support and can reduce the force required to push the foot when in a charged state. It is expected to support more than 7000 steps in actual use. Amin et al. [136] used a wire-driven mechanism and a spring-roll DEA to help patients with hand disabilities recover their motor function as shown in Figure 9d. The torque required for finger extension was determined by analyzing the kinematics and kinetics of the human hand, and a rehabilitation model was designed. The system uses a spring-roll DEA as an intelligent actuator and optimizes its parameters to produce the required output force. In the experiment, a DEA achieved a maximum axial movement of 11.3 mm and a compressive resistance of 4.35 N at an activation voltage of 6.5 kV.

4. Prospects and Challenges

This paper introduces various applications of DEAs in soft robotics. Drawing from the discussions, it is evident that the potential of DEAs in soft robotics is substantial. However, despite the significant progress and impressive results achieved by researchers, the practical implementation of DEAs still faces numerous challenges. This chapter will focus on the key obstacles of DEA technology and explore its future potential applications.

4.1. The Actuation Properties of DE Material

Despite the promising prospects demonstrated by DE materials, they still face a series of challenges in practical applications. Reducing the drive voltage, enhancing the energy and power density, and optimizing the response speed have become urgent problems to be solved.

Modifying DE materials can fundamentally improve their actuation performance and reduce the driving voltage. In early studies, researchers attempted to enhance the dielectric constant of silicone rubber-based DE materials by adding various oxides such as aluminum oxide, barium titanate [137,138], and organically modified montmorillonite (OMMT) [139], as well as piezoelectric materials like lead magnesium niobate-lead titanate (PMN-PT) [140,141] and copper phthalocyanine oligomers (CPO) [142]. For conventional acrylate-based DE materials, in order to reduce the need for pre-stretching, researchers synthesized interpenetrating polymer networks (IPN) to lock the prestrain of the DE films [143,144]. In the modification of other engineering elastomers, the addition of plasticizers (such as dioctyl phthalate, DOP) and high-K ceramics (like TiO₂) [145] has been explored to improve the performance of synthetic elastomers.

Early studies have provided guidance for the modification of dielectric elastomer materials. With advancements in material science and the development of dielectric elastomer fabrication techniques, introducing polar groups [43,66], blending conductive nanomaterials [146,147], and adding high-permittivity ceramic fillers [148,149] have been approved to efficiently enhance the dielectric constant of DE materials. However, the integration of polar groups typically diminishes the dielectric strength as a result of the significant dielectric loss [150,151], and the introduction of conductive or high-permittivity fillers is often beset by the challenge of filler agglomeration, which results in a large stiffness to hinder their actuation strain [150,152]. Sheima et al. [67] selected polymethylvinylsiloxane (PV) as the base material and introduced 2-(methylsulfonyl)-ethanethiol-containing polar sulfonyl groups into the polymer chains via a thiol-ene reaction, forming a modified polysiloxane (PSu). The reaction was controlled to retain some unreacted vinyl groups, and a multifunctional thiol was used as a crosslinking agent to form an elastomer through UV irradiation. The resulting material exhibited a dielectric constant of 18.4 at 10 kHz and 20 °C and demonstrated stable actuation under a low driving voltage of 400 V. Feng et al. [43] developed a polar fluorinated polyacrylate dielectric elastomer, which achieved a high dielectric constant (10.23 at 1 kHz) and ideal modulus (about 0.09 MPa) at low voltage by introducing polar-fluorinated groups and nanodomains formed by the aggregation of long alkyl side chains. This DE exhibited a maximum area strain of 253% at a low driving electric field of 46 MV/m and reached an energy density of 225 J/kg at an electric field of 40 MV/m. Experiments have shown that this DE can be effectively driven at a low voltage of 100 V. The soft crawling robot made of this DE has a movement speed of 20.6 BL/s and can be actuated at a low voltage of 400 V.

Reducing the thickness of the DE film is another effective strategy for lowering the actuation voltage and enlarging the energy output. Zhao et al. [34] fabricated a compact DEA using a spin-coating method to create multilayer sheets, with each layer measuring 25 μm in thickness, which has an actuation voltage below 1000 V and can operate at 200 Hz. Shi et al. [26] developed a processable, high-performance dielectric elastomer (PHDE) with a bimodal network structure and improved electromechanical properties by adjusting crosslinkers and hydrogen bonds. PHDE exhibited a maximum area strain of up to 190% and maintained a strain of more than 110% at 2 Hz without pre-stretching. Peng et al. [88] optimized the dry-stacking process of PHDE and designed novel PDHE ultra-thin film-based multilayer DEAs (PUT-MDEAs). The thickness of a single layer of PHDE film is 10 µm, the ten-layered PUT-MDEAs manifest actuation at 200 V, with the peak areal strain nearing 80% at 800 V. These DEAs offer a blocking force of 0.1 N and an energy density of 22 J/kg at the lower voltage of 200 V, and at the elevated voltage of 700 V, they enhance their blocking force to 0.7 N, along with an increased energy density of 50 J/kg.

4.2. Modeling and Control

Dielectric elastomers are highly nonlinear materials that tend to exhibit variable stiffness and complex dynamic properties when stretched rapidly, such as creep under constant load and hysteresis under cyclic load [153]. Therefore, the mechanical response of DEAs and DEA-driven robots is strongly nonlinear, time-varying, and frequency-varying [154]. In order to achieve effective control of DEAs, the response of DEAs must first be accurately modeled. Once the model is identified, the inverse of the model can be used to develop a feedforward controller, or it can be combined with a feedback loop to develop a more accurate controller.

In previous studies, researchers have attempted to model and explain the electro-mechanical coupling behavior of DEAs using existing physical principles. Suo et al. [155,156,157,158] established and improved a nonlinear force–electric coupling model under the framework of continuum mechanics and thermodynamics, and they analyzed the large deformation, electromechanical instability, viscoelastic behavior, and nonlinear behavior of DEAs. These works not only enhance our understanding of the complex behavior of DEAs but also provide important theoretical support for the subsequent design of new DEAs. Gu et al. [159] developed a constitutive model containing multiple dissipative nonequilibrium mechanisms based on the principle of nonequilibrium thermodynamics and the four-time constant rheological model. The model quantitatively describes the time-dependent response of sheet DEAs in the form of a set of differential equations and the developed corresponding algorithms to solve these equations. The model parameters were adjusted based on experimental data to ensure the consistency between the model predictions and experimental observations. And through the model verification step, it was proven that the proposed model can precisely capture the complex electromechanical responses of DEAs under various voltage loading modes. Jeong et al. [160] used a similar approach and successfully predicted the short-term and long-term behaviors of a spring-roll DEA under dynamic loading, including frequency and time dependencies, verifying the accuracy and applicability of the model. Cao et al. [53] also used a similar method to complete the construction of the dynamic model. And they used a DEA to provide feedback information through changes in self-capacitance and developed a CMAC (Cerebellum Model Articulation Controller) controller based on the cerebellum model to compensate for dynamic uncertainty and provide motion correction.

However, it is still a challenge to fundamentally analyze the complex behavior of DEAs. With the development of phenomenological models and the increase in data-driven methods, there are more options and possibilities for modeling and controlling DEAs. Huang et al. [161] proposed a dynamic model based on a DEA and a tracking control strategy founded on a model predictive controller (MPC). A dynamic model that can simultaneously depict the asymmetric hysteresis, creep, and rate-dependent hysteresis behaviors of the DEA was established. Based on the dynamic model, an inverse compensation controller (ICC) was developed to correct for hysteresis and creep nonlinearities in the DE actuator during tracking control. Additionally, MPC was integrated with ICC to mitigate the impact of model inaccuracies and uncertainties on control precision. Zou et al. [154] proposed a general motion control framework by establishing a control-enabled dynamic model that includes nonlinear electromechanical coupling, mechanical vibration, and rate-dependent viscoelasticity, and they combined it with a state observer and an enhanced exponential arrival law sliding mode controller (EERLSMC). They achieved the accurate trajectory tracking control of DEAs with various configurations, materials, and degrees of freedom. Massenio et al. [162] introduced a reinforcement learning-based algorithm to solve the Hamilton–Jacobi–Bellman (HJB) equation associated with the energy minimization problem. This method uses an adaptive dynamic programming (ADP) algorithm, policy iteration, and value iteration, combined with an actor–critic learning structure, to achieve an approximate solution for the optimal control policy.

Looking back at previous work, the control methods for DEAs are diverse, with various modeling theories enabling accurate model control strategies. As control theory and technology have advanced, model-free control methods have increasingly been applied to DEAs, leading to the development of more intelligent and adaptable control technologies. Furthermore, the integration of intelligent algorithms and multimodal sensors has significantly improved both the precision and autonomy of DEA control. While much of the existing research has focused on single-actuator systems, the expanding applications of DEAs in multiple fields indicate that their role in soft robotics, particularly in automation and the human–robot interaction, will continue to grow. As a result, control methods for DEA-actuated soft robots are gaining more and more attention.

Additionally, it is noteworthy that most efforts in dielectric elastomer control are limited to actuators with simple geometries, while relatively little attention has been paid to the control of soft robots. The control of soft robots is challenging due to their infinite flexible deformation characteristics, so it is difficult to obtain a suitable model. DEA-driven soft robot control is still an active research area and deserves further investigation.

4.3. Reliability

Reliable and trouble-free actuation within the set voltage range is a prerequisite for the commercial application of this actuation technology. A critical factor influencing the reliability of DEAs is the lifespan of the actuator. Elastomers are prone to material degradation under repeated loading, making the study of their fatigue resistance essential. Matysek et al. [163] showed that DE materials undergo significant degradation under cyclic loads. Zakaria et al. [164] showed that pre-stretching can also cause material degradation, resulting in a reduction in the life of the elastomer.

Fatigue is a known issue in elastomers, and the high-voltage operation of DEAs can lead to leakage currents, which in turn generate thermal effects that contribute to material degradation [165]. For instance, when a DEA made from 3M’s VHB is submerged in silicone oil, its breakdown electric field rises from 450 MV/m in air at room temperature to 800 MV/m. This is because DEA in silicone oil is well dissipated and the thermal effect of the material is suppressed [166].

Another critical concern is the consistency and precision of DEAs under varying environmental conditions. Because the glass transition temperature (T_g) is lower than room temperature, the mechanical properties of DEAs can be affected by variations in ambient temperature. These temperature-induced changes require further investigation.

5. Conclusions

The exploration of DEAs within the context of soft robotics has unveiled numerous opportunities and challenges. DEAs, characterized by their high energy density, rapid response, and large deformation capabilities, present a compelling alternative to traditional rigid actuators. Their ability to emulate natural muscle movement has significant implications for advancing soft robotics, particularly in enabling safe interactions with humans and complex environments.

One of the most remarkable advantages of DEAs is their adaptability. The materials employed in DEAs can be tailored to achieve specific performance characteristics, such as optimized stress–strain relationships and actuation speeds, rendering them suitable for diverse applications. However, this inherent adaptability also introduces challenges in standardizing DEAs for large-scale production and ensuring consistent performance across various environmental conditions.

DEAs represent a significant technological advancement in soft robotics, offering capabilities that bring the field closer to realizing robots capable of natural and safe interaction with their environment. Despite the challenges, DEAs hold immense potential across a wide array of applications, ranging from disaster rescue operations to healthcare rehabilitation.

The future of DEAs in soft robotics is promising, with ongoing research efforts directed at overcoming current limitations. Continued advancements in materials science, control theory, and manufacturing techniques are expected to yield DEAs that are increasingly efficient, reliable, and versatile. The integration of DEAs into soft robotic systems will not only broaden the scope of robotic applications but also usher in a new generation of robots designed to work harmoniously alongside humans.

Although transitioning from laboratory research to practical applications presents considerable challenges, the progress achieved thus far underscores the potential of DEAs. As researchers push the boundaries of innovation, the future of soft robotics powered by DEAs is poised to transform the field and create significant opportunities for human–robot collaboration.