Dielectric elastomer actuators are a promising actuation mechanism for soft robots (Gu et al., 2017; Hines et al., 2017). DEAs are capable of high strain deformations at high frequencies (<1 s), reliable blocking forces, and rapid controllability, all of which are valuable qualities for actuation (Choi et al., 2007; Kofod, 2008; Kornbluh et al., 2009; Brochu and Pei, 2010; Anderson et al., 2012; Rosset and Shea, 2012; Carpi, 2016; Madsen et al., 2016a; Naficy et al., 2016). By sandwiching a highly stretchable dielectric layer between two stretchable electrodes, DEAs operate on a simple fundamental principle: charging the electrodes causes an attraction that squeezes the elastomer-based dielectric layer in between them, decreasing the actuator thickness and increasing its area. DEAs can be solid-state or semi-solid (where either the electrodes or the dielectric layer have fluidic elements), and their dielectrics can be made of silicones, acrylic elastomers, and other types of highly deformable, non-conductive materials (Romasanta et al., 2015). Very high electric potentials (in the range of kV) are often necessary to induce significant shape change, despite the dielectric typically being a thin film (<1 mm). For greater details, the reader is recommended these recent reviews (Wang and Qu, 2016; Zhu et al., 2016; Zhang and Serpe, 2017).

The thin nature of DEAs makes them susceptible to defects from physical damage and manufacturing imperfections. These defects are often undetectable until the electric potential is applied (Zurlo et al., 2016), and they can create a continuous conductive bridge or a sudden spark between the electrodes, effectively ending the life of the actuator. Since this defect-induced failure occurs at potentials much lower than required for a general dielectric breakdown, the whole actuator is rendered useless by a single flaw. Identifying this critical issue with DEAs, researchers have begun to search in earnest for effective solutions.

Pneumatic or hydraulic actuators created out of soft elastomeric silicone materials have become ubiquitous in the soft robotics community (Hughes et al., 2016; Lee et al., 2017). These actuators are made by casting the elastomer to create channels that, when inflated, cause the body to bend (or twist) in a manner dependent on the channel design. Their soft, highly deformable elastomer bodies inherently resist damage: absorbing impacts, twisting and compressive loads and more (Martinez et al., 2014), and can even survive being run over with a car (Martinez et al., 2014; Tolley et al., 2014). There is a key problem with this damage resistance: like a balloon, if the actuators pop from overpressure or puncture, they lose all functionality.

Shepherd et al. (2013) presented a potential opportunity to correct this deficiency, wherein they demonstrated a pneumatic actuator that could be punctured with a 14-gauge needle and continued to function. The molded silicone elastomer actuator was reinforced with polyaramid fibers (Kevlar) dispersed into the silicone before curing (Figure 2A). This composite pneumatic arm was punctured with a needle while under positive pressure and maintained its internal pressure, even after the needle was removed (Figure 2A, b–d). Though the reason behind the damage resilience was never fully discovered, the current theory is that the damage resilience stems from a combination of the polyaramid fibers preventing crack propagation (Geethamma et al., 2005) and the silicone’s natural tendency to self-adhere when elastically returning to its original molded shape (after removal of the needle). In another recent example, Wallin et al. (2017) developed a self-healing elastomer-based soft robot, wherein UV cure elastomer resins were used to stereolithographically 3D print hydraulic actuators. The 3D printing process trapped liquid resin in the elastomer; when the robot was punctured, the excess resin was able to seal up the holes if given sufficient UV stimulus (such as direct sunlight).

Terryn et al. (2015, 2017a,b) have taken a different approach to resilient pneumatic actuators by shaping self-healing polymers into inflatable chambers. In their early work, they molded a single, thin-walled cube out of a thermally responsive self-healing polymer and demonstrated its inflation capacity, even after catastrophic failure (Terryn et al., 2015). Although the pneumatic cube was made out of a stiff polymer (when compared to elastomers), it deformed with a 40% increase in side-to-side dimension when inflated (Figure 2B, a). To test its self-healing, the cube was inflated with positive pressure until the chamber popped. To enhance the failure of the actuator, the cube was also cut through with a blade before being heated to instigate self-healing (Figure 2B). After undergoing this puncture and self-healing cycle, the cube inflated the same way it did before it received any damage (both in force produced and displacement achieved). Most recently, Terryn et al. (2017b) demonstrated that by patterning these cubes adjacent to one other, the extended system could be utilized as a soft robotic bending actuator capable of grasping objects. In another recent work, the researchers utilized this same self-healing material to create a pneumatically actuated artificial muscle that can contract over 10% of its length (Terryn et al., 2017a).

Inflatable actuators have seen initial forays into self-healing and damage resilience, and these initial works open possibilities for future development. For example, self-healing inflatable robots open up possibilities in actuator reconfigurability, as the inflatable actuator material can be cut and rearranged as necessary. To achieve such goals, the limitations of the state-of-the-art must be addressed, from long healing times to a need for an intense source of stimulus for self-healing. Since inflatable actuator systems have yet to show any stimuli-independent healing, it will be particularly important in future research to integrate the self-healing stimulus into the soft actuator.

Hydrogels are a type of extremely hydrophilic thermoset polymer matrix that absorb water molecules into the bulk material. By pulling water molecules in between the long polymer chains, a hydrogel’s polymer matrix expands, enabling a change in size many times that of the original structure. This diffusion-limited process has been used repeatedly in soft robots as a potential actuation method (Gerlach and Arndt, 2009; Richter, 2009; Kim et al., 2013). Self-healing hydrogels are their own field of research for materials scientists, as can be seen in recent reviews (Phadke et al., 2012; Taylor and Marc in het Panhuis, 2016). Yet, despite this field of research, hydrogel self-healing has yet to enter the soft robotics community, so this section is mostly forward-looking based on presently developed systems.

Responsive hydrogels have not generally been used in actuators (inflatable or otherwise) like silicones and other inert elastomers. Yuk et al. recently broke this trend in their novel application of hydrogels as an inflatable robotic gripper (Yuk et al., 2017). The hydraulic gripper was built with a highly deformable hydrogel that was simultaneously gentle and robust, capable of trapping a live fish without harming it. Highly stretchable, self-healing hydrogels such as the one recently presented by Jeon et al. (2016) could potentially be used to create these types of hydrogel grippers. Their hydrogel can self-heal very quickly (in under 30 s, given ideal conditions) and can strain over 1,000% its original length (Figure 2C). Similar work has recently been presented by Liu and Li (2017) whose tough, self-healing hydrogel requires slightly more healing time but is also ionically conductive.

There is much to gain by further exploring self-healing hydrogels in soft robots. For example, by adding hygroscopic ions to the hydrogel, the hydrogel could self-hydrate by pulling water out of the ambient environment (Bai et al., 2014; Chen et al., 2014), ensuring that the actuator is always ready to self-heal. The high water concentration in hydrogels (when compared with the humidity of the surrounding air) can enable hydrogel polymers to heal much more rapidly than normal self-healing polymers, as demonstrated by Jeon et al. (2016). However, one constant challenge that hydrogels face is ensuring that they remain sufficiently hydrated. Further advantages gained through the use of conductive hydrogels will be discussed briefly in Section “Shape Memory Materials” in the context of self-healing electronics.

Materials that have been employed in purely functional structural cases are primarily those that can undergo phase changes at low temperatures, such as low melting point alloys (LMPAs) or wax. This is because the low-temperature phase change alters the material stiffness by many orders of magnitude, with relatively small amounts of energy input. These materials are also ideal for self-healing, since they can change to a liquid form that readily flows to fill cracks, holes, or other damage. Once the system has healed the damage, the materials can be cooled again to restore a continuous solid body.

Low melting point alloys have been used in variable stiffness applications in order to provide a stiffening or morphable element in the robot (Nakai et al., 2002; Shan et al., 2013; Shintake et al., 2015), typically display melting points below 70°C and are solid at room temperature. Dr. Dario Floreano’s lab has demonstrated the value of LMPAs as self-healing, variable-stiffness soft robotic components (Figures 3A,B) (Schubert and Floreano, 2013). The researchers started by showing that the solid LMPA encapsulated in an elastomer provides structural stiffness until overloaded, causing the metal to crack and break. When this happens, the load is removed, allowing the encapsulating silicone to restore its original shape and pushing the cracked metal together. By Joule-heating the metal, it melts and reforms a continuous structural system (Figure 3B). Though this process breaks down after several hundred cycles (Figure 3B), it is possible to completely replace the damaged LMPA. Later work advanced this concept by sealing LMPA into a silicone tube that was pre-stretched (see Figure 3C) (Tonazzini et al., 2016). The pre-stretched tube added a constant internal compression to the LMPA, improving the self-healing process when compared to a system using the exact same setup and materials without the pre-stretch.

Polyurethane and silicone foams have been previously used as structural elements in soft robotics, as they are usually easy to deform, compress well, and often can sustain large extensions (Shimoga and Goldenberg, 1992; Lipson, 2014; Yuen et al., 2014, 2016). By impregnating porous foams such as these with low-melting point materials, variable stiffness materials are formed. This structure differs significantly from the above examples of encapsulated LMPA, as the low-melting point materials adhere to the foam (even in a liquid state) using capillary forces and wetting adhesion. Van Meerbeek et al. (2016) mixed LMPA into pre-cured silicone to create an open-celled, foam structure (see Figure 3F). Not only does the LMPA provide stiffening to the foam structure but also enables damage repair by welding the foam block back together with the LMPA post damage (Figure 3F, c). Mirroring the use of LMPAs, Cheng et al. (2014) filled a polyurethane foam lattice with wax which they broke repeatedly though over-compression. This foam–wax structure can self-heal back to nearly full-strength by melting the wax, even after multiple over-compressions. Although not conductive like LMPAs, wax does hold an advantage over LMPA since careful tuning of the molecular structure of the wax will separate out the softening temperature (for physical shape change) from the solid-to-liquid phase change temperature (for self-healing).

Characterization and implementation of soft robotic variable-stiffness, self-healing structures has only just begun. There is much opportunity for research into structural materials that can change shape and heal from mechanical damage by phase changing and reflowing in a soft structure. Finding ways for a soft robotic system to auto-detect damage in a phase-changing structure and initiate the self-repair process is one possible future step that could be taken in this area.

Shape memory materials are a type of material, including both metallic shape memory alloys (SMAs) and shape memory polymers (SMPs), that can be plastically deformed from a pre-programmed state and self-return to their programmed shape upon controlled heating. SMAs and SMPs are common actuator tools that have been used frequently in soft robotics applications (Ionov, 2011; Albu-Schäffer and Bicchi, 2016; Hou, 2016; Rodrigue et al., 2017), but they have only occasionally been used as variable-stiffness structural materials (Manti et al., 2016). Their ability to autonomously restore form after plastic deformation parallels the motivations of self-healing and damage resilience as presented thus far. However, since their common use as actuators does not strictly fall into our definition of damage resilience (the plastic deformation is expected during operation), we will not explore this idea further.

With the common use of SMAs and SMPs as actuators, it would not be difficult to implement them as reinforcing structural elements in damage resilient soft robots (Kirkby et al., 2008; Ferreira et al., 2016). Huang et al. (2010) demonstrated the feasibility of this by embedding a coiled SMA wire inside of a semi-flexible, self-healing polymer (Figures 3D,E). When the polymer structure was stretched beyond its fracture limit, the SMA coil prevented the two segments from separating completely. After failure of the encapsulating polymer, the SMA wire was Joule heated, simultaneously pressing the two fractured halves of the polymer together and heating them so that they could self-heal, thus restoring the polymer flexibility (see Figure 3E, a–f). Xiao et al. (2010) presented a self-healing SMP which could heal its surface after being scratched, which has initiated research into self-healing coatings (Luo and Mather, 2013; Dalmoro et al., 2016; Lutz et al., 2016). SMPs have also been used in two-step self-healing processes where the SMP can close whole cracks using the shape memory effect then self-heal the cracks through internal chemistry (Nji and Li, 2010, 2012; Rodriguez et al., 2011). Such hybrid systems that allow both crack closure and crack healing show much promise, though the biggest challenge will be demonstrating this capacity while simultaneously controlling the soft robot using these same materials. For more details on self-healing and shape memory materials, we highly recommend the first chapter of this book: Li and Meng (2015) or this recent review: Meng and Li (2013).

Gallium-based room temperature liquid metals (eutectic gallium-indium, or eGaIn, and eutectic gallium-indium-tin, or galinstan, alloys) are often used in highly stretchable and flexible soft electronics (Dickey, 2016, 2017). Their unique properties allow for a variety of soft systems to be developed that are not available in solid-metal electronics. First, when encapsulated in channels in a silicone elastomer, the liquid metal flows readily, maintaining continuity as the channel changes geometry due to external extension or compression. Furthermore, when gallium-based liquid metals are exposed to air, the liquid metal forms a stabilizing oxide layer that prevents unwanted flow (Dickey, 2014). These two properties make these alloys ideal for use in damage resilient electronics: the stabilizing oxide prevents the conductive liquid from leaking out any punctures/holes, but the liquid-like behavior allows it to seal circuit breaks.

The effectiveness of room-temperature liquid metals in damage resilient circuitry was recently demonstrated by Li et al. (2016b). In this work, liquid metal channels encapsulated in PDMS silicone were cut and pulled apart, thereby breaking the circuit (Figure 4A, a). When released, the material elastically returned to its original form and the liquid metal reconnected the electrical path (Figure 4A, b). Palleau et al. (2013) used a self-healing, stretchable polymer as an encapsulating body around a room-temperature liquid metal conductive circuit. This combination was used in both flexible self-healing wires and bulk planar materials with liquid-metal channels. The bulk planar material could mechanically and electrically self-heal after being separated into two parts (Figure 4B, a,b). Furthermore, the self-healing enabled reconfigurability in the material, as a 2D electronic pathway could be cut and then reconfigured into a 3D electrically conductive shape (Figure 4B, c,d). This powerful demonstration shows separate self-healing mechanisms for the bulk material and the electrical components, and how to combine the two mechanisms to create a fully self-healing system.

Though self-healing electronics based on liquid metals have been demonstrated in their initial forms, we have yet to see integration of these demonstrations into more complex devices. One potential challenge facing this system is the non-conductive oxide layer that builds up each time the liquid metal is exposed to air (during multiple self-healing events), which results in overall decreasing conductance. However, liquid metal oxidation has also been shown to prevent the liquid metal from leaking during healing events by creating an encapsulating seal at the location of failure (Dickey, 2014). Therefore, while highly advantageous in simple applications, implementation of liquid metal as a self-healing conductor may be limited in applications where stable conductance is necessary, such as resistive sensors.

A second approach to developing fully self-healing electronic systems uses conductive polymer composites formed out of a self-healing polymer matrix and a conductive filler. Appropriate mix ratios can allow for strong electrical conductivity through the bulk material, while maintaining the self-healing properties of the polymer bulk (Tee et al., 2012; Gong et al., 2013). Any cut or puncture in the polymer will self-heal, reconnecting the conductive filler material and rejoining the circuit, much like the material demonstrated by Tee et al. (2012) (Figure 4C). In this work, a polymer composite was demonstrated that could withstand moderate flexing and twisting (withstanding strains of up to 20%), while also being highly conductive and self-healing. The untreated material can withstand significantly higher strains, but when sufficient conductive filler is added (31% vol) to enable high conductivity (10 S/cm), the material’s strain-to-failure is reduced. Alternatively a self-healing elastomer can be coated with a layer of conductive particles, as demonstrated by Cao et al. (2017a). As long as the conductive coating is aligned during the self-healing process, the material regains its conductivity. It is also worth noting that the elastomer presented in this work is able to self-heal in seconds (on the same order of magnitude as hydrogel self-healing, but without the need to be hydrated), which represents a step forward for self-healing elastomeric materials.

Conductive hydrogels that are self-healing at room temperature, robust, and elastically stretchable have begun to emerge in simple composite-polymer sensory applications. A common thread among recent examples is the use of metal ions as a key to both the self-healing chemistry and the conductive element (Darabi et al., 2017; Lei et al., 2017; Liu and Li, 2017; Liu et al., 2017). Though adding metal ions increases the conductivity of the hydrogels, researchers have carefully tuned the concentration: too low and the hydrogel does not conduct, too high and either the flexibility of the hydrogel is reduced or it ceases to self-heal. Cai et al. (2017) took a mixed approach by developing a self-healing, stretchable hydrogel and adding conductive filler (CNTs), much like the composite elastomers previously mentioned. This conductive hydrogel composite can self-heal in 60 s and recover 98% of its electrical conductivity (demonstrated multiple times on the same sample). A third approach to conductivity in self-healing, flexible hydrogels is demonstrated from previous work, wherein the polymer network itself provides the conductivity (Shi et al., 2015).

There are many challenges to overcome as conductive composite soft polymers become more incorporated into damage resilient soft robotic systems. Though the relative electrical stability of many conductive composite polymers has been demonstrated over a few strain cycles, a long-term fatigue cyclic test has yet to prove that these materials do not suffer from electrical drift. The few examples of long-term testing on ionically conductive hydrogels have shown that hydrogel systems can lose water through evaporation during use outside of a solvent environment, and when this happens, their performance drops significantly (Cai et al., 2017). Also, hydrogel composites tend to have higher resistances (greater than 1 kΩ for small samples), limiting their use to mostly strain sensing rather than conductive pathways. For a more comprehensive look at self-healing composite polymers (both bulk composites and conductive composites), readers are referred to these other recent, in-depth reviews/book chapters: Gibson (2010), Benight et al. (2013), Zhu et al. (2014), Ahner et al. (2015), Zhong and Post (2015), Zhang et al. (2017).