1. What soft actuators and artificial muscles are (the short version)

• Devices in which soft materials themselves deform to produce force or motion: instead of motors and cylinders, rubbers, polymers, fibers and gels deform under electric fields, pressure, heat, or water and chemical stimuli. Things that contract, bend or twist like muscle are often called “artificial muscles”; the review by Madden et al. covers dielectric elastomers, ferroelectric polymers, liquid crystal elastomers, shape memory alloys, ionic polymer/metal composites, conducting polymers and carbon nanotubes Sourced [Ref. 8]
• Each approach has a different bottleneck: dielectric elastomers need high voltage and are limited by film breakdown and viscoelasticity; IPMCs and hydrogels by water; pneumatic muscles by compressors and leakage; heat-driven ones by cooling (the speed of response) Sourced [Ref. 1, 10, 13, 11, 22]
• Read a number by its denominator, scope and conditions: energy density is given in kJ/m³ (per volume), J/kg (per mass), kW/kg (power) or N (force), and it also matters whether it is for “the material only” or “the whole device”. Reported lifetimes differ by orders of magnitude, from under 100 cycles to over 300 million Sourced [Ref. 24, 4]

2. A map of drive methods

The approaches can be organized by what serves as the input. Fig. 2 divides them into five: electric field (high voltage), electric field (low voltage, ions), fluid, heat, and water and chemistry, and shows a representative approach in each with the limits confirmed in the sources.

All Sourced (statements of each source). The arrangement is our commentary. Shape memory alloys, shape memory polymers and liquid crystal elastomers (driven by heat or light) are covered in other articles of the series. The values come from different sources, samples and conditions and cannot be ranked.

3. Dielectric elastomer actuators (DEAs) — film, electrodes and voltage

In a DEA, applying a voltage to a thin elastomer film covered on both faces with soft electrodes squeezes the film in thickness and expands it in area through electrostatic force (Fig. 3). The effective pressure from the electrostatic force is commonly written in terms of the relative permittivity ε_r and the field E (= voltage / film thickness) as p = ε₀ ε_r E² (our summary; the review by Zhang et al. also treats the Maxwell stress on the two electrodes in an expression of this kind). Because it scales with the square of the field, a thinner film gives the same field at a lower voltage. Pelrine et al. reported that stretching the film in advance (prestretch) improves performance. Strains were 30 to 40% without prestretch, and up to 117% for silicone and up to 215% for acrylic with it. They state that the strain, pressure and response time of silicone exceeded those of natural muscle and that specific energy densities greatly exceeded those of other field-actuated materials Sourced [Ref. 2, 1].

View the original SVG schematic and exact annotations

The 2025 review by Zhang et al. sets out the challenges of DEAs. Acrylics (such as 3M’s VHB) reach linear strains above 380% and area strains above 2200% (cited values) with large prestretch, but their viscoelasticity and nonlinearity make the response slow; silicones are less viscoelastic and suit fast motion and higher frequencies. DE material needs a low elastic modulus, little viscoelasticity, a high permittivity and a high breakdown field. In one cited example, the breakdown field of VHB rose from 450 MV/m in air to 800 MV/m in silicone oil. Fatigue of the elastomer under repeated loading, and heating from leakage currents under high voltage, can drive degradation. CNT electrodes can recover by themselves after breakdown (self-clearing), which is promising for longer life Sourced [Ref. 1].

Zhang et al. also give device examples. A device reached an area strain above 25% at a drive voltage of 450 V, and a 14% strain at 10 Hz. A DE material had an energy density of up to 244.2 kJ/m³ (cited). Another dielectric elastomer showed an area strain of 253% at 46 MV/m and 225 J/kg at 40 MV/m. A DEA made by coaxial multi-material 3D printing had a life of more than one million cycles at high frequency. A gripper switches between two stable states in 0.17 s. For a flapping-wing robot driven by DEAs, they note that the DE material has sufficient power density to take off but that the overall power density is a challenge Sourced [Ref. 1].

Life changes by orders of magnitude with conditions. Matysek et al. reported that stack DEAs made of Elastosil P7670 (Wacker) could be driven for more than 320 million cycles in a very rough number, regardless of electrical and mechanical driving conditions. The degradation that followed was traced back to the interconnection, and they note that longer lifetimes are conceivable since the connection can be re-established Sourced [Ref. 4].

HASEL couples electrostatic and hydraulic forces by a different route. Acome et al. aimed at soft actuators that are powerful, high-speed, efficient and robust, and showed that they can self-heal repeatedly after dielectric breakdown, with a soft gripper and a self-sensing artificial muscle powering a robotic arm as examples Sourced [Ref. 3].

4. Things driven by low voltage, water, heat and chemistry — IPMCs, hydrogels, CNT yarns, coiled fibers

(1) IPMCs: low voltage, with water as a condition

An IPMC is a sheet of ionic polymer covered with metal electrodes (the most common technique is chemically plating a commercially available polymer membrane with Pt or Au); it bends at low voltage and also generates a voltage when bent. The bending voltage is typically below 5 V, with a 2.5 V example, and the frequencies it can actuate run from DC to about 100 Hz. The response depends on the sample geometry and ambient humidity, and the measured voltage also depends on the hydration level and excitation frequency Sourced [Ref. 10]. Shahinpoor and Kim summarize the fundamentals of IPMCs, including ways of developing high-force-density IPMCs [Ref. 9].

(2) Hydrogels: water is both the power and the constraint

A hydrogel is a crosslinked network containing 80–90 wt% water that swells or shrinks with temperature, light, pH, ionic strength, magnetism and pressure. Above the lower critical solution temperature (LCST) the volume and shape change, which is used in actuators. The review by Banerjee et al. says that electro-responsive hydrogels have a slow response and develop high stress that causes fatigue, that low-polymer-content gels are weak and dehydrate easily, and that thin coatings of polydimethylsiloxane and similar materials can restrain dehydration. Remedies for the slowness include making the gel smaller and combining pNIPAM with single-walled carbon nanotubes to speed up the thermal response five-fold Sourced [Ref. 13].

(3) CNT yarns: twisting in an electrolyte

Foroughi et al. reported that an electrolyte-filled twist-spun carbon nanotube yarn, much thinner than a human hair, works as a torsional artificial muscle in a three-electrode electrochemical system, giving a reversible 15,000° rotation and 590 revolutions per minute. They explain that contraction along the length and torsional rotation occur simultaneously when the yarn volume increases through electrochemical double-layer charge injection (the hydrostatic mechanism of muscular hydrostats) Sourced [Ref. 6]. Lima et al. reported hybrid CNT yarns powered by electricity, chemistry and light [Ref. 7].

(4) Coiled fibers: a large stroke from inexpensive fibers, driven by heat

Haines et al. reported that inexpensive high-strength polymer fibers used for fishing line and sewing thread, twisted into coils, become fast, scalable, nonhysteretic, long-life tensile and torsional muscles. Extremely twisted fibers contracted by 49%, lifted loads over 100 times heavier than a human muscle of the same length and weight, and generated 5.3 kilowatts of mechanical work per kilogram of muscle weight (“similar to that produced by a jet engine”). Woven textiles whose porosity changes with temperature and actuating window shutters were also demonstrated Sourced [Ref. 5]. As the limits of other artificial muscles they name high cost, low cycle life, hysteresis and low efficiency, and they present this approach as inexpensive, nonhysteretic and long-life.

5. Pneumatic artificial muscles

The basis of a pneumatic artificial muscle is that highly compressed air is pumped into and out of a rubber tube restrained by a coaxial woven sheath, converting radial expansion and contraction into axial force. The review by Su et al. describes pneumatic soft robots as lightweight, efficient, non-polluting and robust in harsh environments (they can operate without ferromagnetic or electronic components under strong radiation, electromagnetic interference, dust and impact). As an example, a pneumatic artificial muscle with a 20 mm inner diameter is said to provide 1500 N of pulling force, enough for practical loads Sourced [Ref. 11].

The review by Walker et al., on the other hand, says that general compressors use a lot of electricity, so alternative power sources based on chemical reactions or combustible fuels are being examined, and that portable power supplies, circuit boards and drive components are issues for the future [Ref. 12]. Su et al. point out that thin geometries need completely bubble-free construction to avoid leakage or failure [Ref. 11].

6. Reading “muscle-like” — denominator, scope, time and life

(1) Denominator and scope: the same “density” means different things

The performance of artificial muscles is described by energy per volume (kJ/m³), energy per mass (J/kg), power per mass (kW/kg), force (N) and so on. Fig. 6 shows the denominator and scope to which each value in this article belongs. A review of shape memory alloy actuators reports their energy density at 5000–25,000 kJ/m³, far above the 40–70 kJ/m³ of human skeletal muscle [Ref. 22]. Zhang et al. give the energy density of a DE material as up to 244.2 kJ/m³ (cited) and, for another material, 225 J/kg at 40 MV/m [Ref. 1]. These are values for the alloy or the material itself, not for the whole device including heating and cooling, the power supply and the bias spring.

(2) Time: what does “fast” count?

Speed of response also counts different times. A DEA gripper switching between two stable states takes 0.17 s (a switching time), and a device giving a 14% strain at 10 Hz has a cycle of 0.1 s (a period) [Ref. 1]. IPMCs actuate from DC to about 100 Hz [Ref. 10]. For an LCE the examples are a response time of 20 s and a 162° bend in 30 s [Ref. 24]. A shape memory alloy actuator usually takes several seconds per motion, with cooling setting the speed [Ref. 22]. Electro-responsive hydrogels are described as slow [Ref. 13]. Fig. 7 lays out the reports that give a number.

(3) Life: reports that differ by orders of magnitude

Lifetimes differ by orders of magnitude between approaches and devices. A stack DEA reached more than 320 million cycles (until the interconnection degraded) [Ref. 4], and a coaxially 3D-printed DEA more than one million cycles [Ref. 1]. For LCEs, lightly crosslinked ones are prone to fatigue, many run for fewer than 100 cycles, and an electrically driven example was stable over 50 [Ref. 24]. For shape memory alloy (nitinol), an FDA regulatory science tool states that low-cycle fatigue below about 10 million cycles is associated with cyclic phase transformation [Ref. 23]. HASEL can heal repeatedly after dielectric breakdown [Ref. 3]. The definition of the “one cycle” being counted (strain, temperature, load) differs, so a comparison of orders of magnitude is only a guide to what has been confirmed Sourced.

7. Limits, and what we could not confirm

(1) Coverage of numbers

Values of output density, strain and life for IPMCs, HASEL and pneumatic artificial muscles are limited in the abstracts and texts we confirmed, and we do not give them. We have not read the performance summary table in the review by Madden et al. We have not checked the original papers for the values cited by the reviews (380%, 2200%, 244.2 kJ/m³, one million cycles, fewer than 100 cycles and so on).

(2) How the sources were read

We read the text of the DEA review by Zhang et al., the pneumatic reviews by Su et al. and Walker et al., the hydrogel review by Banerjee et al., and the IPMC review by MohdIsa et al. For Pelrine et al., Acome et al., Haines et al., Foroughi et al., Lima et al., Matysek et al., Madden et al. and Shahinpoor and Kim we checked the abstracts on publisher or bibliographic database pages.

(3) The text of the standards

We have not read the text of IEC 60243-1 and -2, IEC 62631-2-1 and -2-3, ISO 6943, ISO 37, JIS C 2110-1, C 2139-2-1, K 6270 or K 6251. The ASTM International website states that AI use is prohibited, so we did not open any ASTM site; ASTM standards are listed only with the number and what the test measures, as confirmed in a manufacturer’s datasheet and a testing laboratory’s page.

(4) Dedicated standards

A search of the ISO catalogue (10 October 2026) found no standard for “electroactive polymer”; the results for “artificial muscle” and “soft robot” were unrelated standards, such as detection of larvae in meat by artificial digestion and fracture toughness of ceramics. A JISC title search found no JIS for “artificial muscle”, “soft robot”, “electroactive” or “dielectric elastomer” (the Japanese terms). We could not confirm any IEC or IEEE standard that defines performance tests for artificial muscles or soft actuators.

(5) Medicine, safety and regulation

We do not cover the safety of soft actuators in medical or wearable use, electrical safety (handling high voltage) or regulation. No product or company is ranked.

(6) Other approaches

Shape memory alloys, shape memory polymers and liquid crystal elastomers are treated in other articles of the series, and piezoelectrics and ferroelectrics, conductive polymers and smart textiles in others again. Here they appear only as references for comparison.

8. How to read evaluations and standards

The evaluation of soft actuators divides into (a) properties of the material (modulus, tensile properties, relative permittivity, breakdown field), (b) the output of a device (displacement and strain, force, work, power), (c) dynamic behavior (response time, frequency response) and (d) life (number of cycles, degradation). For (a), general standards for rubber and insulating materials can be used. We could not confirm standards that directly specify (b), (c) and (d); each paper uses its own test apparatus and definitions.

Table of corresponding standards

JIS titles are our translations from the Japanese.

Numbers, titles and editions were confirmed as follows: ISO on the ISO catalogue pages; IEC on the IEC Webstore; JIS in the JISC database; ASTM through a manufacturer’s datasheet (Loctite, 1995), a testing laboratory page (Intertek) and the FDA database; we did not open any ASTM site [Ref. 14, 15, 16, 17, 25, 18, 19, 28, 20, 21, 26]. The ASTM editions are those given in those sources and are not necessarily the latest. The arrangement of the table is our commentary.

The number, title, edition and date of each standard are Sourced (confirmed on the official pages of the standards bodies, the Japanese Industrial Standards Committee, a manufacturer and a testing laboratory). Which standard goes in which cell is our own arrangement (Our calculation). We found no revision or draft in progress, so no row falls under Not yet confirmed.

Cautions when reading standards

1. Standards for insulating materials test the material, not the actuator. IEC 60243-1 gives short-time electric strength at power frequency (48 to 62 Hz) [Ref. 14]. When a DEA runs on slowly varying voltage or direct current, consider the additional requirements for direct voltage (IEC 60243-2) [Ref. 15] and the difference from how the film is actually used (our commentary).
2. Read a datasheet value with its thickness and frequency. In the Loctite silicone example, the ASTM D149 dielectric strength is for a “4 mm film”, and the ASTM D150 permittivity is listed by frequency, about 3.0 at 100 Hz and about 3.1 at 1 MHz [Ref. 20]. For the tens-of-micrometre films used in DEAs and a 4 mm thick film, the values need not be the same even for the same material (our commentary).
3. A tension-fatigue standard for rubber is not a life test under an applied field. ISO 6943 concerns fatigue under repeated tension with little temperature rise [Ref. 18]. The life of a DEA involves the field, heating by leakage current, the electrodes and degradation of the interconnection [Ref. 1, 4].
4. With no dedicated standard, read the test conditions in the paper. The “output”, “response time” and “life” of artificial muscles are defined differently in each paper. Check the denominator and scope in Fig. 6 and the definition of time in Fig. 7.

9. Glossary

10. References (primary sources)

1. Zhang et al. “A Review of the Applications and Challenges of Dielectric Elastomer Actuators in Soft Robotics”, Machines 13, 101 (2025), DOI 10.3390/machines13020101 (open access; we read the text) — doi.org/10.3390/machines13020101
2. Pelrine R. et al. “High-Speed Electrically Actuated Elastomers with Strain Greater Than 100%”, Science 287, 836–839 (2000), DOI 10.1126/science.287.5454.836 — doi.org/10.1126/science.287.5454.836
3. Acome E. et al. “Hydraulically amplified self-healing electrostatic actuators with muscle-like performance”, Science 359, 61–65 (2018), DOI 10.1126/science.aao6139 — doi.org/10.1126/science.aao6139
4. Matysek M. et al. “Lifetime investigation of dielectric elastomer stack actuators”, IEEE Trans. Dielectr. Electr. Insul. 18, 89–96 (2011), DOI 10.1109/TDEI.2011.5704497 — doi.org/10.1109/TDEI.2011.5704497
5. Haines C. S. et al. “Artificial Muscles from Fishing Line and Sewing Thread”, Science 343, 868–872 (2014), DOI 10.1126/science.1246906 — doi.org/10.1126/science.1246906
6. Foroughi J. et al. “Torsional Carbon Nanotube Artificial Muscles”, Science 334, 494–497 (2011), DOI 10.1126/science.1211220 — doi.org/10.1126/science.1211220
7. Lima M. D. et al. “Electrically, Chemically, and Photonically Powered Torsional and Tensile Actuation of Hybrid Carbon Nanotube Yarn Muscles”, Science 338, 928–932 (2012), DOI 10.1126/science.1226762 — doi.org/10.1126/science.1226762
8. Madden J. D. W. et al. “Artificial Muscle Technology: Physical Principles and Naval Prospects”, IEEE J. Ocean. Eng. 29, 706–728 (2004), DOI 10.1109/JOE.2004.833135 — doi.org/10.1109/JOE.2004.833135
9. Shahinpoor M., Kim K. J. “Ionic polymer-metal composites: I. Fundamentals”, Smart Mater. Struct. 10, 819–833 (2001), DOI 10.1088/0964-1726/10/4/327 — doi.org/10.1088/0964-1726/10/4/327
10. MohdIsa W. H. et al. “Sensing and Self-Sensing Actuation Methods for Ionic Polymer–Metal Composite (IPMC): A Review”, Sensors 19, 3967 (2019), DOI 10.3390/s19183967 (open access) — doi.org/10.3390/s19183967
11. Su H. et al. “Pneumatic Soft Robots: Challenges and Benefits”, Actuators 11, 92 (2022), DOI 10.3390/act11030092 (open access) — doi.org/10.3390/act11030092
12. Walker J. et al. “Soft Robotics: A Review of Recent Developments of Pneumatic Soft Actuators”, Actuators 9, 3 (2020), DOI 10.3390/act9010003 (open access) — doi.org/10.3390/act9010003
13. Banerjee H. et al. “Hydrogel Actuators and Sensors for Biomedical Soft Robots: Brief Overview with Impending Challenges”, Biomimetics 3, 15 (2018), DOI 10.3390/biomimetics3030015 (open access) — doi.org/10.3390/biomimetics3030015
14. IEC “IEC 60243-1:2013 Electric strength of insulating materials – Test methods – Part 1: Tests at power frequencies” (IEC Webstore) — webstore.iec.ch/en/publication/1101
15. IEC “IEC 60243-2:2013 Electric strength of insulating materials – Test methods – Part 2: Additional requirements for tests using direct voltage” (IEC Webstore) — webstore.iec.ch/en/publication/1102
16. IEC “IEC 62631-2-1:2018 Dielectric and resistive properties of solid insulating materials – Part 2-1: Relative permittivity and dissipation factor – Technical Frequencies (0,1 Hz – 10 MHz) – AC Methods” (IEC Webstore) — webstore.iec.ch/en/publication/34510
17. IEC “IEC 62631-2-3:2024 Dielectric and resistive properties of solid insulating materials – Part 2-3: Relative permittivity and dissipation factor – Contact electrode method for insulating films – AC methods” (IEC Webstore) — webstore.iec.ch/en/publication/70208
18. ISO “ISO 6943:2017 Rubber, vulcanized – Determination of tension fatigue” (ISO catalogue page) — www.iso.org/standard/72790.html
19. Japanese Industrial Standards Committee (JISC) JISC JIS list (C 2110: dielectric breakdown strength). JIS C 2110-1, C 2139-2-1, K 6270 and K 6251 were confirmed (number, title, dates) on the JISC list and detail pages, and title searches for artificial muscle, soft robot, electroactive and dielectric elastomer (in Japanese) returned no JIS — www.jisc.go.jp/app/jis/general/GnrJISNumberNameSearchList?show&jisS…
20. Loctite (Henkel) Technical Data Sheet, Product 5145, Worldwide Version, October 1995 (datasheet of a silicone product; used to confirm the ASTM D150 and D149 numbers and the 4 mm film thickness statement) — www.farnell.com/datasheets/16681.pdf
21. Intertek “Dielectric Strength ASTM D149, IEC 60243” (testing laboratory technical page; used to confirm the number and what the test measures) — www.intertek.com/polymers-plastics/testlopedia/dielectric-strength-…
22. Romančík J. et al. “Shape Memory Alloy Actuators in Robotics”, Actuators 15, 162 (2026), DOI 10.3390/act15030162 (open access) — doi.org/10.3390/act15030162
23. U.S. FDA, Center for Devices and Radiological Health (CDRH) “Strategy to Estimate Low to High Cycle Fatigue Transition of Nitinol for Fatigue to Fracture Test Planning” (Catalog of Regulatory Science Tools) — cdrh-rst.fda.gov/strategy-estimate-low-high-cycle-fatigue-transitio…
24. Hasan M., Liu Y. “4D-Printed Liquid Crystal Elastomers: Printing Strategies, Actuation Mechanisms, and Emerging Applications”, J. Compos. Sci. 9, 633 (2025), DOI 10.3390/jcs9110633 (open access; we read the text) — doi.org/10.3390/jcs9110633
25. ISO “ISO 37:2024 Rubber, vulcanized or thermoplastic — Determination of tensile stress-strain properties” (ISO catalogue page) — www.iso.org/standard/86892.html
26. U.S. Food and Drug Administration (FDA) “Recognized Consensus Standards: Medical Devices” database, search results for titles containing tension testing (including ASTM D412; page last updated 28 September 2026) — www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfStandards/results.cfm?…
27. Staszczak M. et al. “Characterization of Polyurethane Shape Memory Polymer and Determination of Shape Fixity and Shape Recovery in Subsequent Thermomechanical Cycles”, Polymers 14, 4775 (2022), DOI 10.3390/polym14214775 (open access; we read the full text) — doi.org/10.3390/polym14214775
28. Japanese Industrial Standards Committee (JISC) JIS search (by number or title). JIS Z 2255, C 2501, K 7121, K 7122, R 1611, H 7801, R 3257, R 1703-1, H 7101 and others were confirmed (number, title, dates of establishment and revision) on the JIS list and detail pages — www.jisc.go.jp/app/jis/general/GnrJISSearch.html

11. Claim-to-source audit

Last updated: 10 October 2026. Sources are limited to primary material (peer-reviewed papers, the databases of standards bodies and the Japanese Industrial Standards Committee, manufacturers’ datasheets and a testing laboratory’s page). We have not read the text of the standards. Because the article includes organization of structure and readings of materials design, these are marked as “commentary” and kept apart from sourced facts. We could not confirm, and therefore do not state, the output density, strain and life of IPMCs, HASEL and pneumatic muscles, any standard that directly specifies the performance of artificial muscles and soft actuators, the text and test conditions of ASTM standards, or medicine, safety and regulation. The strain, force, energy density, speed and lifetime values are reports for specific papers and devices and cannot be generalized to other devices or products. We give no medical advice and no ranking of any product or company. Figs. 2 to 4, 6 and 7 are vector drawings and Figs. 1 and 5 are AI-generated images; none of them shows an actual micrograph or a real product.