Shape Memory Alloy Actuators in Robotics

Abstract

Shape memory alloys (SMAs) are materials that, when used as actuators, can generate deformation and force that can be used to perform mechanical work. This actuation capability is driven by temperature variation, which induces a reversible phase transformation between martensite (at low temperature) and austenite (at high temperature). Owing to their advantages, SMAs are widely applied as actuators and, in certain applications, can be more suitable than other actuation technologies. A thorough understanding of SMA actuator characteristics is therefore essential for their effective implementation in practical applications. This article provides an overview of the most important properties of SMA actuators. In addition, it reviews the application potential of SMA actuators in robotics. Based on the survey of the literature, perspectives for further research and development in this field are also presented.

1. Introduction

Since ancient times, humans have sought to make life easier. At first, they facilitated their work using various tools and later through machines and robots. Such devices, intended to assist in performing tasks, must incorporate actuators that convert energy (electrical, pneumatic, hydraulic, etc.) into mechanical work. This mechanical work is then used to position other parts of a machine, for example for lifting, opening and closing, adjustment, or other technological operations. An actuator—often referred to as a drive—functions in a manner analogous to a human muscle: based on control signals, it generates physical motion or force. Actuators are widely employed in automation and robotics, as well as in the automotive industry, aerospace engineering, industrial equipment, medical devices, household appliances, and many other fields [1,2,3,4,5].

According to their structural configuration, actuators can provide either rotary or linear output motion. Based on the underlying physical operating principle, several types of conventional actuators can be distinguished:

• electromagnetic actuators;
• electric motor—driven actuators;
• pneumatic actuators;
• hydraulic actuators.

In addition, several non-conventional actuator types exist, including:

• piezoelectric actuators;
• magnetostrictive actuators;
• electrostatic actuators;
• thermal actuators based on shape memory alloys (SMAs);
• wax-based thermal actuators.

Research and development of all actuator types is continuously advancing; consequently, the performance characteristics of individual actuator classes (

Table 1. Actuator characteristics [7,8,9,10,11,12,13,14,15,16,17,18,19,20].

Actuator	Max. Stress (MPa)	Max. Strain (%)	Efficiency (%)	Power Density (W/cm3)
Shape memory alloy (NiTi) [10,11,12]	200–600	6–8	1–10	10–100
Piezoceramic (PZT) [13]	30–100	0.1–0.15	30–70	1–10
Electromagnetic (solenoid/voice coil) [14,15]	1–10	5–50	20–60	0.5–5
Pneumatics [20]	0.5–5	>100	10–30	0.05–1
Hydraulics [20]	10–50	>100	40–70	1–10
Magnetostrictive (Terfenol-D) [16]	70–150	0.1–0.2	20–40	5–20
Electrostatic (MEMS) [7,17,18]	0.01–1	10–100	50–90	0.01–0.1
Thermal wax (Paraffin) [19]	1–10	10–20	<1	0.01–0.1

) are steadily improving, and new actuator types are also emerging [6].

For specific applications, actuators are selected according to the performance expected under the given operating conditions. However, an appropriate selection requires detailed knowledge of the characteristics and application limitations of the individual actuator types [2,21]. Prior to choosing a suitable actuator, several key selection criteria should be addressed:

• What type of motion must be implemented (linear or rotary)?
• What working stroke/rotation and what output force/torque are required?
• What positioning accuracy and operating speed are required?
• What level of efficiency and overall effectiveness is required?
• What additional requirements must be satisfied (e.g., dimensions, mass, control strategy, power-supply constraints, electrical current consumption, required ingress protection rating (IPXX), capability to operate in explosive atmospheres and environments with elevated fire risk, permissible operating temperature and humidity ranges, resistance to chemical agents, minimum service life, cost, etc.)?

A major challenge is the efficiency and effectiveness of actuators relative to their physical size. This issue is particularly critical in the miniaturization of certain actuator types, where reduced dimensions lead to a loss of efficiency and performance [8,22]. The dimensional differences between the macro- and microscales result in distinct physical regimes, affecting the dominant phenomena, object motion, and the relative scaling of system performance across these scales. Accordingly, microphysics can be defined as the natural science concerned with studying motion, structural properties of objects, and thermal, optical, electrical, and magnetic phenomena that change or become dominant during actuator miniaturization. For instance, attractive surface forces are dominant at the microscale compared to gravitational forces, which dominate at the macroscale.

To explain the impact of miniaturization on system efficiency, scaling laws are frequently employed. The performance of micro-actuators is strongly dependent on scaling effects. Based on scaling analysis, it is possible to identify a suitable actuation principle (drive mechanism). Scaling laws also represent a powerful tool for the design of micromechatronic systems that incorporate micromechanisms, energy delivery components, and related subsystems. Particular attention is therefore devoted to micro-actuators, which are becoming key components for achieving high overall system efficiency [23,24,25,26,27,28].

Smart actuators represent a popular category of actuators. “Smart” actuators incorporate a certain level of sensing capability by integrating sensors for force, displacement, temperature, pressure, strain, and/or other quantities, thereby enabling a degree of self-monitoring. Such smart actuators typically provide improved accuracy and reliability, which supports broader and more versatile use. Owing to feedback and the integrated elements and functionalities listed above, they can exhibit a differentiated and quasi-intelligent response to varying operating conditions and stimuli. The application scope of these actuators is relatively wide and extends into the medical domain and the automotive industry. Smart actuators play an important role in applications such as smart automotive suspension systems (smart shock absorbers), intelligent braking systems, fuel injection systems, noise and vibration control systems, and others [29,30,31].

To properly select an actuator, it is necessary to fully understand the capabilities of available actuators and their limitations. Therefore, this article aims to review actuators made of shape memory alloys, called SMA-Shape Memory Alloys actuators.

Publications focused on research into shape memory alloy (SMA) actuators were surveyed in the citation databases Web of Science (WoS), Scopus, and IEEE Xplore. The existence of shape memory materials has been known for approximately 90 years, and the first scientific description is attributed to the Swedish physicist Arne Ölander (1932) [32]. A major breakthrough in the field occurred in 1962 with the discovery of the NiTi alloy (Nitinol) at the U.S. Naval Ordnance Laboratory (Buehler and Wang) [33,34], which exhibited properties highly attractive for practical applications. Since this discovery, research on shape memory alloys has intensified. A marked increase in interest in SMA research and applications began around 1983 and has continued to rise rapidly to the present (Figure 1).

The publication analysis was performed with deduplication across the individual databases; therefore, the reported counts represent unique publications that appear at least once in any of the above-mentioned databases.

An informative perspective is provided by the percentage distribution of publications on shape memory alloys across individual research fields, which characterizes the level of researchers’ interest in the application of shape memory alloy actuators (Figure 2).

Another informative overview is the evolution of the number of publications on shape memory alloy (SMA) actuators in the field of robotics (Figure 3). This trend (Figure 3) essentially mirrors the overall growth in publication activity related to shape memory alloys (Figure 1).

Research in robotics can also be subdivided into specific subdomains. This analysis (Figure 4) shows that the most substantial expansion has occurred in research on grippers, motion control, and soft robots. The chart (Figure 4) also indicates potential trends for future research and development in robotics.

1.1. Data Collection, Inclusion, Deduplication and Classification Procedure

1.1.1. Data Collection and Search Strategy

A systematic literature review was conducted using the Web of Science (WOS Core Collection), Scopus and IEEE Xplore databases. The search was conducted on 21 January 2026. The following search strings were used:

(“shape memory alloy” OR “SMA”) AND (“actuator”).

Searchable fields:

• WOS: Title, Abstract, Author Keywords, Keywords Plus
• Scopus: Title, Abstract, Keywords
• IEEE Xplore: Document Title, Abstract, Author Keywords, IEEE Terms

Filters:

• Document type: Article, Conference Paper, Review
• Language: English
• Time limit: no time limit (from the oldest available records to the end of 2025)

Export of found records:

Records were exported as full bibliographic data including DOI, title, authors, publication year, abstract and keywords.

WOS: 5310 records

Scopus: 17,289 records

IEEE Xplore: 3853 records

Total 26,452 records

1.1.2. Exclusion of Irrelevant Publication Records

The inclusion and exclusion criteria for the publication record were chosen with respect to the focus of this article.

Inclusion criteria: A publication was included if:

• It contained an explicit mention of SMA materials (e.g., “shape memory alloy”, “SMA”),
• It also dealt with their use as an actuator (not just material research),
• It had a content overlap with technical or robotic applications.

Exclusion criteria: Publications were excluded if:

• focused exclusively on material characterization without an actuation application,
• related exclusively to superelasticity without an actuation use,
• without an available abstract (if relevance could not be verified),
• were clearly irrelevant (e.g., biomedical stents without an actuation function).

Relevance was assessed based on a combination of title, abstract, and keywords.

1.1.3. The Process of Deduplication of Publication Records Between WOS, Scopus and IEEE Xplore Databases

Since the same publications were found in multiple databases, a multi-stage deduplication process was applied.

Duplicate identification rules:

• Primary rule—DOI match: If two records had identical DOIs → they were considered duplicates. One record was kept (preferably with full abstract).
• Secondary rule—title + year + first author match: Used in case of missing DOIs. Publication names were normalized (converted to lowercase; removed punctuation; removed extra spaces).
• In case of conflicts, the record with full abstract, available DOI, richer metadata was kept.

1.1.4. Publication Record Classification Process

After deduplication, the publications were classified into two levels:

• Application area (robotics, medicine, aerospace, MEMS, civil construction, industry, etc.)
• Thematic focus in robotics (mobile robots, microrobotics, grippers, manipulators, exoskeletons, control, etc.)

Classification categories were entered manually based on the analysis of the frequency of occurrence of keywords in the deduplicated dataset of publications, which were then sorted into these application areas. Based on this sorting, publications were then automatically divided into these areas (

Table 2. Primary Application Areas and Representative Classification Keywords.

Primary Area	Representative Keywords/Expressions
Robotics	robot, robotic, manipulator, gripper, locomotion, mobile robot, exoskeleton, micromachine, continuum robot
Automotive	automotive actuator, vehicle system, active aerodynamics, engine control, transmission actuator, adaptive suspension
Medicine/Biomedical Engineering/Medical Applications	biomedical, medical device, surgical tool, stent, catheter, implant, minimally invasive, prosthesis
Aerospace/Space Applications	aerospace, space structure, satellite, deployable structure, morphing wing, space mechanism
Micromachine/MEMS/Microsystems	MEMS, microactuator, microsystem, micromirror, microgripper, microdevice
Civil Structures/Vibration Control	vibration control, seismic damping, structural control, smart structure, active vibration
Mechanical Engineering/Industrial/Mechanical Applications	valve, pump, industrial actuator, automation system, positioning system, precision mechanism
Energetics, Energy Systems	energy harvesting, thermomechanical conversion, heat recovery, energy conversion
Electronics/Optics/RF	RF switch, tunable antenna, optical switch, adaptive optics, reconfigurable device
Materials Research (excluded if no actuation context)	phase transformation, microstructure, fatigue behavior, martensitic transformation

).

Each article was classified into only one subcategory (Table 2) to avoid double-counting in the analysis of shares and trends. Ambiguous publications were classified manually. This entire classification process was solved iteratively until all publications were classified into an area. A random sample of approximately 10% of publications was manually reviewed to verify the consistency of the assignment. Ambiguous cases were corrected to maintain the logical consistency of the taxonomy.

Similarly, robotics subdomains were created for the field of robotics. Subdomains were manually created according to the frequency of occurrence of keywords, and then individual keywords were assigned as sorting characters for individual areas (

Table 3. Robotics Subdomains and Representative Classification Keywords.

Robotics Subdomain	Representative Keywords/Expressions
Robotic Grippers and End-Effectors	gripper, prehensile, end-effector, grasping, compliant finger
Manipulators and Robotic Arms	robotic arm, manipulator, joint actuation, articulated mechanism
Ground Mobile Robots (UGVs)/Wheeled/Tracked/Legged/Crawled	mobile robot, wheeled robot, tracked robot, locomotion platform
Flying Robots/UAVs	UAV, unmanned aerial vehicle, drone, aerial robot, flapping wing robot
In-Pipe/Inspection Robots	in-pipe robot, pipeline inspection, pipe crawler
Snake/Continuum Robots	snake robot, continuum robot, hyper-redundant robot
Microrobotics/Micromachines	microrobot, micromachine, micro-swimmer, untethered microdevice
Wearable Robotics/Exoskeletons	exoskeleton, wearable robot, assistive device
Soft Robotics	soft robot, soft actuator, compliant robot, elastomeric structure
Motion control and Modeling for Robotics	motion control, position control, hysteresis compensation, modeling, feedback control, trajectory tracking
Other Robots	bio-inspired robot, hybrid robot, reconfigurable robot, experimental robotic platform

). Subsequently, publications were automatically divided into individual groups. 20% of the publications were manually reviewed for consistency in assignment to a specific robotics field.

1.1.5. Pseudocode of Publication Record Classification Process

The pseudocode below outlines the workflow used to compile, clean, deduplicate, and classify records retrieved from WoS, Scopus, and IEEE Xplore. All exported entries were first pooled into a single dataset. To make records comparable across databases, key metadata were standardized, with particular emphasis on DOI formatting and title normalization; the first author was also extracted to support duplicate detection when DOI information was missing.

Relevance screening was then performed using the combined Title + Abstract + Keywords fields, and only publications with a clear link to SMA-based actuation were retained. Duplicate records were removed in two steps. First, entries sharing the same DOI were treated as duplicates. Second, for records without DOI, duplicates were identified using a composite key based on normalized title, publication year, and first author.

After deduplication, each publication was assigned to a primary application area based on terminology in the title, abstract, and keywords. For records classified as robotics-related, a second-level categorization was applied to capture more specific robotics subtopics. In this step, we also accounted for common terminology differences (e.g., papers using “micromachine” instead of “robot”), when the context clearly referred to mobility or other functions characteristic of robotic systems.

The final output is a deduplicated and systematically classified dataset with application areas and, where applicable, robotics subtopics.

Pseudocode:

INPUT: WOS_records, Scopus_records, IEEE_records STEP 1: Merge all records R ← WOS_records ∪ Scopus_records ∪ IEEE_records STEP 2: Normalize metadata for each record r in R: r.doi_norm ← normalize_DOI(r.DOI) r.title_norm ← normalize_title(r.Title) r.first_auth ← first_author(r.Authors) r.key_noDOI ← (r.title_norm + r.Year + r.first_auth) STEP 3: Apply inclusion criteria (Title + Abstract + Keywords) R ← keep r where has_SMA_terms(r) AND has_actuator_terms(r) STEP 4: Deduplicate R ← deduplicate_by_DOI(R) R ← deduplicate_by_key_noDOI(R)  // (title_norm + year + first_author) for records without DOI STEP 5: Classify application area for each record r in R: r.area ← assign_primary_area(r.Title, r.Abstract, r.Keywords) STEP 6: Classify robotics topic (only for robotics) for each record r in R where r.area == “Robotics”: r.topic ← assign_robotics_topic(r.Title, r.Abstract, r.Keywords) // treat “micromachine” as robotics if context indicates robot mobility/function OUTPUT: Deduplicated + classified dataset R (area, topic)

The entire process was implemented using scripted metadata processing, with deduplication and classification rules being deterministic and applied consistently to all records.

1.2. Organization of Rest of the Paper

The rest of the paper is organized as follows. Section 2 describes the key properties and phenomena characteristic of shape memory alloy (SMA) actuators. It also presents the most common structural configurations and setups of SMA actuators as well as methods of their activation and deactivation. Section 3 reviews the application potential of SMA actuators, with an emphasis on robotics-related domains. Section 4 provides a discussion of the reported application outcomes and outlines perspectives for future research and development in this field.

2. Materials and Methods

Shape memory alloys (SMAs) are metallic materials composed of an alloy of two or more metals that “remember” their original shape and, after deformation and subsequent heating to the transformation temperature, recover this initial shape. The original shape is defined by appropriate heat treatment [12,23].

The crystallographic structure of SMA materials exhibits two phases: martensite (at low temperature) and austenite (at high temperature) (Figure 5). When comparing the force-to-volume ratio of an SMA actuator with that of other actuator types, SMAs can provide exceptionally high force output.

In the martensitic phase at low temperatures, this material can be readily deformed; upon subsequent heating to the transformation temperature, it recovers its original shape. This phenomenon has been known since the 1960s, and several alloy systems are currently known to exhibit the shape memory effect. The most widely recognized is the nickel–titanium (NiTi) alloy, commonly referred to commercially as Nitinol. In addition to NiTi, other alloys with similar behavior exist (e.g., gold–cadmium, copper–aluminum, and copper–aluminum–nickel, among others), and research in this field continues to expand (

Table 4. Shape memory alloy materials [10,11,35,36,37,38,39].

Alloy	Weight Fraction (%)	Transformation Range (°C)
Ag-Cd [10,11,35,36,37]	Ag ~75%, Cd ~25%	−50 to +100
Au-Cd [10,11,35,36,37]	Au ~50%, Cd ~50%	−100 to +100
Cu–Al–Ni [10,11,35,36,37,38]	Cu ~82–88%, Al ~11–14%, Ni ~3–5%	−140 to +100
Cu–Zn–Al [10,11,35,36,37,38]	Cu ~68–80%, Zn ~15–30%, Al ~3–8%	−200 to +100
Ni–Ti (Nitinol) [10,11,35,36,37]	Ni ~55%, Ti ~45%	−100 to +110
Fe–Mn–Si [10,11,35,36,37,39]	Fe ~65–70%, Mn ~25–30%, Si ~5–6%	−200 to +150
Ni–Ti–Cu [10,11,35,36,37]	Ni ~45%, Ti ~45%, Cu ~10%	−50 to +100
Ni–Al [10,35,37]	Ni ~62%, Al ~38%	+100 to +250
Co–Ni–Al [10,35,37]	Co ~38–40%, Ni ~33–35%, Al ~25–27%	−50 to +200

).

Actuators made from SMA materials are available in the form of wires, helical springs, or strips (Figure 6). Although helical springs provide a larger stroke, this is achieved at the expense of the force generated by the actuator [40].

A change in the crystalline structure of an SMA actuator—and thus the generation of force and deformation—can be induced by internal heating via an electric current, commonly referred to as Joule heating. However, heating can also be provided by an external heat source. Cooling of SMA actuators is most commonly achieved through heat dissipation to the surrounding environment at room temperature. In applications requiring faster actuator dynamics, forced cooling is necessary, using methods such as forced air convection, passive heat sinks, cooling liquids, Peltier elements, or combinations of these approaches.

SMA actuators are designed to perform mechanical work by undergoing a temperature change from room temperature to the transformation temperature (Figure 7). In actuator implementations, SMA materials are typically combined with mechanical structures so that the actuator’s shape, dimensions, or stiffness changes as the SMA temperature varies. The actuation response of SMA actuators exhibits pronounced hysteresis (Figure 7), which poses a significant challenge, particularly for actuator control [41,42,43,44].

SMA actuators offer several advantages:

• They are compact and lightweight, providing a high power-to-mass ratio and high energy density. The energy density is reported to be in the range of 5000–25,000 kJ·m⁻³, whereas human skeletal muscle exhibits only 40–70 kJ·m⁻³ [12,45].
• SMA actuator activation can be achieved using a power source capable of regulating a constant electric current.
• SMA actuators operate quietly.
• A key advantage is their biocompatibility, which makes them well-suited for applications in biomedical engineering.

SMA actuators also exhibit several disadvantages:

• A very limited stroke (typically up to ~5% of the active length).
• A slow response time (one motion cycle usually takes several seconds, depending on actuator dimensions and the heating and cooling approach).
• One of the most significant limitations is the long response time, which is governed by the heating and, in particular, the cooling rate. Cooling is typically the primary bottleneck; therefore, dedicated cooling methods must be considered.
• Pronounced nonlinearity and hysteresis in the relationship between stroke and excitation current complicate position control of SMA actuators. If only end-to-end motion between two stroke limits is required, control is generally straightforward; however, difficulties arise when accurate control to intermediate positions is needed.

The hysteresis of SMA actuators (Figure 7) results from the behavior of these alloys during cyclic heating and cooling, due to thermal losses and internal friction during the phase transformation in the SMA material [46].

The structured comparison in

Table 5. Comparative Performance of SMA Actuator Forms in Robotics.

SMA Form	Typical Recoverable Strain	Typical Force/Stress	Typical Frequency	Cooling
Straight wire [47,48]	3–4% (≤5% safe cyclic)	150–400 MPa recovery stress	0.1–1 Hz	Natural convection/forced air
Coiled wire (spring) [49,50]	5–20% geometric stroke	5–50 N (geometry dependent)	0.05–0.5 Hz	Passive/limited active cooling
SMA bundle (multi-wire) [49,51]	3–4% per wire	50–200 N (bundle dependent)	0.05–0.2 Hz	Sequential activation
Foil/thin strip [52,53]	2–4%	Thickness dependent	1–5 Hz (micro-scale)	Improved surface cooling
Tube/rod [47]	2–4%	High axial force	<0.5 Hz	High thermal mass
Functionalized coil/fiber [54]	10–40% geometric stroke	0.5–2 N (micro-scale)	0.5–1 Hz	Enhanced heat transfer

Notes: Performance ranges are representative values reported in cited literature. Frequency strongly depends on thermal conditions and geometry.

highlights that the performance of SMA actuators strongly depends on geometric configuration rather than solely on intrinsic material properties. While straight wires provide high recovery stress and predictable cyclic durability at moderate strain levels (≤3–4%), their usable stroke is limited and thermally constrained. In contrast, coiled configurations amplify displacement at the expense of mechanical bandwidth and fatigue life.

Multi-wire bundles enable higher force output suitable for wearable and assistive robotic systems; however, thermal accumulation significantly reduces achievable cycle frequency. Thin foils and strips benefit from improved surface-to-volume ratio, allowing higher operational frequencies, particularly in micro-scale applications.

Advanced functionalized coils and fiber-based actuators demonstrate promising improvements in heat dissipation and frequency response, although long-term fatigue behaviour remains insufficiently characterized.

Overall, the comparison confirms that actuator selection in robotic design represents a trade-off among stroke, force density, thermal dynamics, and durability. Table 5 directly supports the limitations discussion by linking reported dynamic constraints (e.g., sub-Hz operation in many wearable systems) to actuator geometry and cooling strategy.

2.1. Actuator Design Configuration—Setup of Actuator

From a practical perspective, there are one-way SMA actuators (one-way shape memory effect, OWSME) that generate force and displacement only in a single direction. For example, an SMA wire contracts upon heating and thus produces mechanical work. The return to the original shape is possible only through an external restoring force, also called a bias force (e.g., gravitational preload, spring preload, or an antagonistic arrangement of two one-way SMA actuators) (Figure 8). The preload force must be sufficiently high to enable the reverse transformation from the austenitic phase back to the martensitic phase. Otherwise, the original shape will not be recovered even after the SMA material is no longer heated.

The simplest method of generating a preload force is gravitational preloading (Figure 8); however, in this arrangement the actuator must remain in a fixed orientation. This can be a drawback for mobile applications. In addition, the applied weight represents extra mass that must be transported in mobile systems. If the device configuration allows, it is advantageous when the gravitational preload simultaneously serves as the external load force.

Using a spring preload is generally more practical in terms of overall size and mass of an SMA-actuated system. This approach can operate in any orientation. Nevertheless, it must be considered that the useful output force available for the actuator’s mechanical work is reduced by the spring force.

An antagonistic configuration employing two one-way SMA actuators (Figure 8) is also a feasible solution; however, the net output force is likewise reduced by the force required to deform the opposing, antagonistically connected SMA actuator [55,56].

The torsional hinge SMA actuator (Figure 9) consists of an SMA wire (SMA) trained to a 180° bending angle and a second, superelastic wire (SE) that provides the bias force required for returning the mechanism to its initial state during cooling of the SMA element. A gripper based on these torsional actuators can grasp fragile objects of various shapes and sizes (Figure 9) without causing damage—an outcome that is often challenging for conventional grippers, which typically require integrated force sensors to detect and regulate the grasping force [57].

In addition to the one-way shape memory effect (OWSME), the two-way shape memory effect (TWSME) also exists, in which a shape memory alloy can “remember” two working shapes (one at low temperature and another at high temperature) without the need for an external preload force. In TWSME alloys, a second shape is “programmed” into the material, which the actuator adopts upon cooling. As a result, the actuator assumes one shape when heated and a different shape when cooled. Consequently, this actuator type can realize a full stroke without any additional biasing forces. However, the practical use of TWSME-based actuators is less common due to the requirements associated with “programming” and, in particular, because the achievable actuation amplitude is typically reduced (approximately by half) compared with one-way configurations [53].

Another phenomenon observed in shape memory alloys is pseudoelasticity, in which the alloy recovers its previous shape after mechanical loading when operated within the temperature range where phase transformation between austenite and martensite can occur, without the need for thermal activation. For actuator applications, this effect is generally not suitable [58].

Authors often address the limitations associated with the actuator’s small stroke and the need for preloading by employing a pulley system (Figure 10) together with either a tensile spring preload or an antagonistic configuration of two SMA actuators [59,60].

Another approach to utilizing SMA material is the implementation of a displacement amplification structure (Figure 11). The proposed actuator consists of supporting discs separated by preloading springs that keep the SMA wires under tension. When the SMA wires are thermally activated, the supporting discs move closer to each other. The arrangement of the SMA wires provides motion amplification of the SMA contraction, which would otherwise be too small for many practical applications [61,62].

Some authors integrate SMA materials into composite actuator structures (Figure 12), often referring to them as “smart composites”. Such systems are used, for example, as shape-morphing composites in aerospace, automotive, biomedical, and other application domains. These actuator architectures are commonly implemented in two configurations (Figure 12): the fully embedded smart morphing composite actuator and the hybrid embedded smart morphing composite actuator [63].

A fully embedded smart morphing composite actuator (Figure 12) is a structure in which an SMA wire actuator is completely embedded within a compliant host structure, which may be made of a polymer, a fiber-reinforced polymer, an elastic metallic structure, or another elastically deformable material. In contrast, a hybrid embedded smart morphing composite actuator (Figure 12) places the SMA wire outside the composite material, enabling the formation of a deformable structure with a targeted, application-specific deformation profile [63].

With respect to the resulting actuation mode, SMA smart morphing composites can be realized in various structural forms (for example bending, twisting, contracting, or expanding smart morphing composites [63]).

Fully embedded SMA composite structures can be supplemented with low-stiffness polymer scaffold matrices to induce out-of-plane deformation of the composite structure in a desired direction (Figure 13). The composite structure may consist of multiple scaffold layers and multiple SMA elements [23].

2.2. Termination and Fixation of SMA Actuators

Reliable termination and fixation of SMA wire actuators are essential for practical implementation.

Reliable fixation (and simultaneous electrical contacting) of SMA wire actuators is most commonly achieved using purely mechanical methods, such as screw-clamp terminations or crimped sleeves/end-fittings, analogous to terminations used for electrical conductors (Figure 14). These approaches are preferred because they do not introduce a thermally affected zone into the functional NiTi material and are scalable to practical actuator assemblies. In fact, surveys of contacting solutions report that, in industrialized applications, crimp or splice connectors are predominantly used for SMA wire integration, while other approaches are often limited by manufacturability and repeatability [64]. Crimping is also frequently described as a common state-of-the-art method for combined mechanical and electrical connection of SMA wires, particularly when compact solutions are needed [65].

However, standard clamping/crimping can suffer from stress concentrations and variability of contact resistance, motivating alternative form-fit concepts. For example, laser-based processing of SMA wire ends has been proposed to create robust form-fit terminations that remain functional under cyclic loading [66]. Such approaches aim to improve repeatability and cyclic robustness while maintaining a joining process that minimizes adverse thermal influence on the active section of the SMA wire.

In contrast, soldering and welding are generally less suitable for SMA actuator terminations because the required heat input can alter microstructure and phase-transformation behavior, thereby degrading functional properties (shape memory effect, superelastic response) and reducing fatigue life. Comprehensive reviews on NiTi joining emphasize that the preservation of functional properties after joining is a central challenge, due to the sensitivity of NiTi to thermal cycles and the formation of brittle phases/precipitates depending on the process [67]. Experimental studies on NiTi wire joining similarly highlight that high-temperature interactions can negatively affect material characteristics and complicate the production of reliable joints [68]. Although brazing/soldering routes can be engineered to reduce thermal impact, they still require careful process control and may introduce reaction layers and microstructural changes within the joint region, which can influence the attainable functional strain and failure behavior [69].

Adhesive bonding is therefore typically considered a supplementary method (e.g., for strain relief, insulation, or fixation of a mechanical termination), rather than a primary load-bearing termination technique, because cyclic actuation induces repeated mechanical loading of the bond line, which can compromise long-term reliability—especially in compact, thermally cycled SMA systems [67].

2.3. Excitation of SMA Actuators

Shape memory alloy (SMA) actuators are activated by a thermal stimulus that induces a phase transformation of the material from martensite to austenite, thereby producing mechanical deformation. The heat required for this transformation can be supplied to the SMA in two principal ways: by direct external heating or by indirect heating through the passage of an electric current.

The first approach is direct (external) heating, which can be achieved via heat convection, thermal conduction, or thermal radiation. Direct heating may be realized by flowing hot air or another gas. Immersion in a heated liquid can also be used for external heating. Another option is a solid-state contact heater (e.g., a resistive heater). In principle, heating by radiation from an infrared heat source is also possible. Because SMA actuators are metallic, induction heating using a high-frequency magnetic field is likewise feasible, where eddy currents heat the SMA material. This method enables galvanic isolation of the actuator from the electrical circuit; however, it is typically characterized by lower efficiency and slower dynamic response.

The second—and in practice the most widely used—approach is indirect heating by passing an electric current directly through the SMA element (also referred to as resistive heating), in which the material is heated due to Joule heating. This method offers higher energy efficiency, faster response, and simpler integration into control systems; therefore, it is employed in the majority of engineering applications of SMA actuators.

The Joule heat generated by the passage of electric current through the material, i.e., during resistive heating of an SMA actuator, can be approximated as:

$$Q_J=I^2\cdot R\cdot t,$$

(1)

where I is the electric current flowing through the material, R is the electrical resistance of the material, and t is time.

This heat induces the martensite–austenite phase transformation (actuator contraction), and after the current is reduced, the actuator cools and returns to its original dimensions and shape.

Electrical current excitation of an SMA actuator is also advantageous from a control perspective, because the actuator temperature—and thus its actuation state—can be regulated by adjusting the magnitude of the current flowing through the SMA element. Manufacturers typically specify a recommended current level at which the material undergoes the phase transformation and, consequently, the actuator deforms. Exceeding the specified excitation current can lead to a substantial temperature rise, potentially causing permanent material degradation or a significant reduction in the service life of the SMA actuator.

The actuation response of an SMA actuator can be expressed as stress–strain–temperature ($\sigma$-$\epsilon$-T) constitutive simplified model by the following equation [70]:

$${\displaystyle \frac{dR}{R}}={\pi }_e\cdot d\sigma +K_{\epsilon }\cdot d\epsilon +{\alpha }_{RT}\cdot dT,$$

(2)

where R is resistivity, ${\pi }_e$—piezoelectric coefficient, $\sigma$—stress, $K_{\epsilon }$—coefficient of shape sensitivity, $\epsilon$—strain (deformation), ${\alpha }_{RT}$—coefficient of thermal expansion, $T$—temperature.

Comprehensive models of SMA actuators have been the subject of numerous publications. SMA electro–thermo–mechanical models are relatively complex and are used to describe the behavior of SMA actuators. Brinson’s constitutive model has been addressed in several studies [67,68,69,70]:

$$\sigma -{\sigma }_0=E\left(\xi \right)\epsilon -E\left({\xi }_0\right){\epsilon }_0+\Omega \left(\xi \right){\xi }_S-\Omega \left(\xi \right){\xi }_{S0}+\Theta \left(T-T_0\right),$$

(3)

where subscript 0 denotes the initial conditions, Ω(ξ) is a scalar function named “transformation tensor” by Brinson that can be expressed as a function of the elastic modulus E(ξ) and the transformation strain ε_L, as Ω(ξ) = −ε_L E(ξ), and Θ is related to the thermal expansion coefficient for the SMA material [71,72,73,74].

The hysteretic relationship between temperature (or input current) and strain was modeled as a Preisach model using a discretized Preisach operator [75]:

$$\epsilon \left(t\right)={\displaystyle \sum _{i=1}^N{w}_i{\gamma }_i\left[T\left(t\right)\right]},$$

(4)

where γ_i are elementary relay operators and w_i are experimentally identified weights. In the discretized Preisach formulation, N denotes the number of elementary relay operators used to approximate the continuous hysteresis operator, thereby determining the resolution of the hysteresis representation [48].

In some studies, this model is further extended to include a dynamic mechanical model [72].

When an actuator is excited by a step change in the electrical current passing through it, its temperature rises with a certain time delay (Figure 15). After the maximum deformation is reached, the excitation current is switched off abruptly, and as the actuator cools gradually, it returns to its original shape and dimensions.

The activation time under electrical current excitation can also be expressed as the ratio of the energy required to heat the actuator to the electrical energy consumed [76]:

$$t_a={\displaystyle \frac{E_h}{P_{el}}},$$

(5)

where the energy required to heat the actuator E_h and the electrical energy consumed for actuator activation, P_el is electrical power dissipated in the actuator (Joule heating) and it can be expressed as:

$$E_h=c_v\cdot m\cdot \Delta T=c_v\cdot {\rho }_M\cdot S_{NiTi}\cdot L\cdot \Delta T,$$

(6)

$$P_{el}=I^2\cdot R={\displaystyle \frac{I^2\cdot {\rho }_E\cdot L}{S_{NiTi}}},$$

(7)

where c_v is specific heat constant, ΔT is temperature difference in crystallization phases, m is actuator weight, ${\rho }_M$ is density of the actuator, ${\rho }_E$ is electrical resistivity of SMA actuator, $S_{NiTi}$ is cross section area of actuator, L is actuator length, I is heating electric current.

Note that the electrical resistivity ${\rho }_E$ is not constant during actuation. As shown in Figure 16, ${\rho }_E\left(T\right)$ exhibits a pronounced hysteresis associated with the martensite↔austenite transformation; therefore, both $R\left(T\right)={\rho }_E\left(T\right)L/S_{NiTi}$ and the Joule heating power $P_{el}=I^2\cdot R\left(T\right)$ vary over the thermal cycle even under constant-current driving. This behavior can be linked to the phase fraction, since a common first-order approximation expresses the effective resistivity as a mixture of phase resistivities:

$${\rho }_E\left(T\right)\approx {\xi }_M\left(T\right){\rho }_M+\left(1-{\xi }_M\left(T\right)\right){\rho }_A,$$

(8)

where ${\xi }_M\left(T\right)\in \left[0,1\right]$ is the martensite fraction (and ${\xi }_A=1-{\xi }_M\left(T\right)$ is the austenite fraction). ${\rho }_M$ is electrical resistivity of martensite fraction and ${\rho }_A$ is electrical resistivity of austenite fraction. Consequently, the hysteresis loop in Figure 16 reflects the hysteretic evolution of ${\xi }_M\left(T\right)$ during heating and cooling [77,78,79,80].

The activation time can then be expressed as:

$$t_a={\displaystyle \frac{A_{NiTi}\cdot S_{NiTi}^2\cdot \Delta T}{I^2}},$$

(9)

where A_NiTi is material constant for SMA actuator.

This relationship indicates that the actuator activation time depends on the material properties, the actuator cross-sectional area, and the applied current. Therefore, the actuator length does not have a significant influence on the activation time. The deactivation time is more difficult to describe because it depends on the cooling method.

During electrical current excitation of an SMA actuator, the change in actuator temperature is accompanied by a change in the electrical resistance of the SMA material (Figure 16). Therefore, the current supply must be capable of stabilizing the current to prevent a decrease in the flowing current due to variations in the actuator’s electrical resistance. For control purposes, constant-current excitation is often combined with pulse-width modulation (PWM).

The schematic in Figure 17 illustrates the operating principle of an analog current driver intended for shape memory alloy (SMA) actuators. The control signal is generated by the microcontroller (MCU) as a PWM waveform, which is combined in the PWM modulator block with the reference voltage VREF to produce an analog control signal proportional to the desired current. This signal is fed into the Current Sense Amplifier block, where—within a closed feedback loop—it is compared with a voltage proportional to the actual current flowing through the SMA actuator. Based on this feedback, the output current is automatically regulated to maintain a constant value, independent of changes in the SMA actuator resistance caused by heating.

The constant current through the SMA actuator ensures controlled heating by Joule losses and thus triggers the actuator’s mechanical deformation. At the same time, the voltage across the SMA actuator is routed to the OP AMP block, where it is amplified and conditioned to the ADC_Voltage level suitable for measurement by the MCU’s ADC. From the measured voltage, the SMA actuator resistance can be determined, which is directly related to its temperature and phase state. This information can be utilized for closed-loop actuator control without the need for external position or temperature sensors [81].

A similar solution is presented in Figure 18, which illustrates a linear constant-current driving circuit for a shape memory alloy actuator based on an operational amplifier and a power transistor operating in the linear region. The operational amplifier compares a reference voltage defining the desired actuation current with the voltage drop across a current-sensing resistor R_S placed in series with the SMA wire. Through negative feedback, the output of the operational amplifier controls the gate or base of the power transistor, thereby regulating the current flowing through the SMA actuator.

As a result (Figure 18), the actuator current is determined primarily by the reference voltage and the sensing resistor and remains largely independent of temperature-dependent variations in the electrical resistance of the SMA material. The circuit further enables measurement of the voltage across the SMA wire, allowing indirect estimation of its electrical resistance, which can be employed for self-sensing or closed-loop control strategies. Overall, the presented driving scheme ensures stable and reproducible Joule heating of the SMA actuator, which is essential for reliable actuation performance [82].

A constant-current mode is important because SMA wires exhibit resistance variations during heating; maintaining a constant current therefore ensures a controllable output temperature and, consequently, precise motion. The described analog solution, which operates without a digital microprocessor, provides low latency and high accuracy without the need for PWM or software-based control. The approach is adaptable to different types of SMA actuators—i.e., it can accommodate various electrical resistances, lengths, and wire geometries—because the feedback loop regulates the current according to the actual operating conditions [83].

2.4. Cooling of SMA Actuators

Cooling of an SMA actuator represents a critical phase of its operating cycle, because the return of the material to the martensitic phase and its original shape requires the temperature to drop below the transformation temperature Ms (or Mf). Unlike heating, which can be controlled relatively easily by adjusting the supplied power, the cooling process is strongly influenced by ambient conditions and by the actuator’s design.

Under passive cooling—most commonly achieved by natural convection to the surrounding air—the return to the martensitic phase is slow and limits the maximum operating frequency of the actuator. The thermal inertia of the SMA and the small temperature gradient between the actuator and the environment lead to long cooling time constants, which are particularly pronounced for wires and strips with larger cross-sections. The time dependence of the SMA actuator temperature can be approximately described by the following relation:

$$m\cdot c_p\cdot {\displaystyle \frac{dT}{dt}}=-{\epsilon }_S\cdot \sigma \cdot S\cdot \left(T^4-T_S^4\right)$$

(10)

where:

T(t)—temperature of SMA wire actuator;

T_S—ambient temperature;

ε_S—emissivity of the SMA wire actuator surface;

σ—Stefan-Boltzmann constant (5.670374419·10⁻⁸ W·m⁻²·K⁻⁴);

m—mass of the SMA actuator material;

S—surface area of the SMA wire actuator excluding the end faces (π·D·L);

L—length of SMA wire actuator;

D—diameter of SMA wire actuator.

Passive cooling is mechanically simple and energy-efficient; however, it does not allow fast and repeatable actuation dynamics. For this reason, applications with higher demands on speed or cyclic operation employ active cooling, for example by forced air (liquid) convection or by using heat sinks [84,85]. Active cooling significantly reduces the time required to return to the martensitic phase, thereby increasing the actuator operating frequency; however, this comes at the expense of higher energy consumption, increased design complexity, and often larger overall system dimensions.

Special attention must also be paid to the thermal hysteresis of SMAs (Figure 7), which causes the martensitic transformation temperatures during cooling (Ms, Mf) to be lower than the austenitic transformation temperatures during heating (As, Af). As a consequence, cooling may be incomplete if the temperature does not decrease sufficiently, leading to incomplete shape recovery or reduced available stroke in the subsequent cycle. This phenomenon directly affects positioning accuracy and the long-term stability of the actuator.

Cooling is the primary limiting factor for the dynamic performance of SMA actuators. In most practical applications, the cooling rate determines the maximum operating frequency, efficiency, and motion repeatability. Therefore, successful SMA actuator design requires a compromise between cooling speed, energy consumption, and the overall mechanical simplicity of the system.

The concept of a wet shape memory alloy (SMA) actuator (Figure 19) [84] is inspired by biological systems, where fluids are used for thermoregulation, material transport, and also for muscle activation. Several authors have investigated wet SMA actuators that are, for example, fully embedded in fluid-filled vessels, using Joule (resistive) heating for activation and a cold working fluid for cooling.

Other approaches employ separate reservoirs for hot and cold fluids; alternating between these fluids can provide improved SMA actuator dynamics. However, the size and mass of such actuators—including the associated heating and cooling subsystems—are typically large, making these concepts attractive mainly for stationary applications. For mobile systems, they are currently of limited practical potential. Nevertheless, there are applications where sufficient working fluid is readily available (e.g., automatic control of a valve in an automotive cooling circuit), and the ability to operate the actuator without electrical power can be advantageous. Other wet shape memory alloy actuator solutions have also been published [86,87].

2.5. Control Strategies of SMA Actuators

Although SMA actuators offer high force density and structural simplicity, their practical deployment in robotic systems is strongly constrained by nonlinear thermomechanical behavior, rate-dependent hysteresis, and thermal inertia. Consequently, control strategy and sensing architecture become central design parameters rather than secondary implementation details.

Across the reviewed literature, four dominant control paradigms can be identified:

➢

Open-loop current control (constant current or PWM-based)

➢

Closed-loop position/force control

➢

Model-based control with hysteresis compensation

➢

Sensorless (self-sensing) control based on resistance estimation

2.5.1. Open-Loop and Current-Based Control

Early implementations frequently employed constant current driving or PWM-based current regulation due to hardware simplicity. PWM-based control enables efficient power delivery while maintaining average current levels; however, performance is typically limited by:

➢

thermal time constants (0.1–2 s depending on wire diameter),

➢

nonlinear resistance–temperature relation,

➢

absence of hysteresis compensation.

Reported closed-loop bandwidths in such configurations rarely exceed 1–3 Hz in wire-based robotic actuators [73,88].

2.5.2. Closed-Loop Control with External Sensing

For robotic manipulators, grippers, and continuum robots, closed-loop feedback is typically implemented using:

➢

position sensors (encoders, LVDTs, flex sensors),

➢

force sensors (load cells),

➢

temperature sensors (thermocouples).

Typical loop frequencies reported in robotic systems range from 50–500 Hz sampling, and the effective mechanical bandwidth is 1–10 Hz.

Tracking errors depend strongly on hysteresis compensation strategy, with reported steady-state position errors between <1% (model-based control) and 5–10% (simple PID without compensation).

Closed-loop PID remains the most common strategy in practical robotic prototypes [47,89,90].

2.5.3. Hysteresis Compensation and Model-Based Control

Given the pronounced nonlinear hysteresis of SMA materials, several works adopt:

➢

Preisach-based models

➢

Bouc–Wen models

➢

Prandtl–Ishlinskii models

➢

phenomenological thermomechanical models

Model-based feedforward + feedback control significantly improves tracking performance, often reducing hysteresis-induced errors by 50–80% compared to pure PID [91,92,93].

Reported improvements include:

➢

reduced overshoot,

➢

faster settling times,

➢

improved repeatability,

➢

enhanced robustness to ambient temperature variations.

However, real-time implementation requires computationally efficient model reduction.

2.5.4. Self-Sensing (Resistance-Based Sensing)

Resistance-based self-sensing exploits the intrinsic correlation between electrical resistance and phase transformation state.

Shape memory alloy (SMA) actuators can provide an inherent self-sensing functionality, because the electrical resistance (or resistivity) of the SMA element varies during Joule heating and cooling (Figure 16) and correlates with the phase transformation state and the resulting strain/displacement (Figure 20). Therefore, the actuator itself can be used as a feedback transducer (Figure 21) by estimating resistance from measured voltage and current, which may reduce or eliminate external sensors and support miniaturization and integration in robotic mechanisms [77,94,95].

Advantages:

➢

eliminates external position sensors,

➢

reduces system mass and wiring complexity.

Limitations:

➢

resistance–strain relationship is nonlinear and temperature-dependent,

➢

sensitive to noise and supply fluctuations.

Self-sensing implementations in robotics report:

➢

sampling frequencies up to 1 kHz,

➢

position estimation errors typically 2–5% after calibration [95,96,97].

In practice, the resistance–displacement relationship is affected by hysteresis, thermal gradients, stress/preload and cycling history; thus, many implementations rely on

➢

separate heating/cooling mappings,

➢

hysteresis compensation, or

➢

calibration under a fixed preload/duty-cycle regime [94,95].

Early experimental work demonstrated that displacement can be inferred from the voltage drop (resistance change) of SMA wire actuators with high correlation, highlighting the feasibility of sensorless position estimation [98]. More recent approaches combine self-sensing with robust identification/compensation and even data-driven calibration (e.g., machine learning) to improve accuracy in compact layouts where voltage probe placement and parasitic resistances become limiting factors [99,100].

Self-sensing has been demonstrated in robot-relevant mechanisms, including antagonistic SMA drives (Figure 22) for precise servo control [94], compliant SMA-driven grippers [101], and flexible surgical instruments where actuator deflection (“shape-sensing”) is estimated directly from electrical resistance for minimally invasive robotics [102].

2.5.5. Thermal Management Considerations

Thermal dynamics fundamentally limit response speed. Cooling strategies reported include:

➢

passive convection [103,104],

➢

forced air cooling [103,104,105],

➢

heat sinks [56,105,106,107],

➢

antagonistic actuation for faster recovery [56,94,108,109].

Thermal management directly affects achievable bandwidth and duty cycle, often representing the main limiting factor in high-frequency robotic applications [56,103,104,110,111,112,113].

2.5.6. Control Strategies in the Context of SMA Actuator Applications in Robotics

The comparative overview in

Table 6. Control strategies for SMA actuators in robotics.

Application	Control Method	Sensing Strategy	Driver Type	Loop Frequency	Reported Performance
SMA gripper [88]	Constant current + PWM	None (open-loop)	PWM current driver	-	Slow response (<1 Hz), qualitative control
Continuum robot [89]	PID position control	External position sensor	Linear current driver	100 Hz	<5% tracking error
Robotic manipulator [92]	Model-based + PID	Encoder	Current-controlled PWM	200 Hz	50–70% hysteresis reduction
Micro-robot [95]	Self-sensing control	Resistance-based	PWM	1 kHz sampling	2–5% position estimation error
Soft robotic actuator [91]	Preisach + feedback	Force sensor	Current driver	50 Hz	Improved repeatability, reduced overshoot
Exoskeleton joint [96]	Adaptive control	Resistance + temperature	PWM	500 Hz	Stable control under load variation

highlights that the control strategy is not merely a secondary implementation detail but a decisive factor shaping the practical viability of SMA actuators in robotic systems.

First, in subdomains such as robotic grippers, continuum robots, and soft robotics, where structural compliance is intrinsic to the mechanical design, classical PID control with current regulation remains the most frequently reported approach. In these systems, the required mechanical bandwidth is typically low (1–5 Hz), and the primary objective is stable position or force regulation rather than high-speed tracking. Consequently, simple closed-loop control with external sensing is often sufficient, particularly in laboratory prototypes [88,89,91,92].

In contrast, in wearable robotics and exoskeleton applications, higher demands are placed on responsiveness, repeatability, and disturbance rejection under varying load conditions. Here, adaptive or model-based control strategies are more commonly reported, often combined with PWM current drivers operating at sampling frequencies in the range of 200–500 Hz. The literature indicates that hysteresis compensation can reduce steady-state tracking error by more than 50% compared to uncompensated PID control, which is critical for human–robot interaction safety and comfort [96].

A different control paradigm appears in microrobotics and MEMS-scale systems, where mass and space constraints favor resistance-based self-sensing approaches. In these applications, eliminating external sensors simplifies integration and reduces system complexity. However, the achievable accuracy (typically 2–5% position estimation error) remains strongly dependent on calibration and thermal stability. As observed in the reviewed works, sensorless control is particularly attractive for untethered or miniaturized robotic systems but remains sensitive to environmental variations [95].

Across all robotics subdomains analyzed in this review, thermal dynamics emerge as the fundamental limiting factor. The reported mechanical bandwidth rarely exceeds 5–10 Hz for wire-based SMA actuators without active cooling. Therefore, the suitability of SMA technology for a given robotic architecture is tightly coupled to the dynamic requirements of the task. Applications involving quasi-static grasping, shape adaptation, or low-frequency morphing are well aligned with SMA characteristics, whereas high-speed cyclic motion remains challenging without advanced thermal management.

Another notable trend revealed by this review is the increasing shift from purely current-based open-loop strategies toward integrated model-based approaches. As the field matures, the emphasis is moving from proof-of-concept prototypes toward performance-optimized robotic systems. This evolution mirrors the broader transition observed in smart material actuation, where constitutive modeling and control integration are becoming essential for engineering-grade solutions [47,48,114,115].

Thus, the control architecture cannot be considered independently of mechanical design and thermal management. Instead, SMA actuator performance in robotics must be evaluated as a coupled electro-thermo-mechanical control problem, where actuation, sensing, and heat transfer are inherently interconnected.

3. Application Potential of Shape Memory Alloy Actuators in Robotics

These materials are widely used in automotive applications, aerospace engineering, civil engineering, and biomedical devices. The present article further focuses on applications in robotics. Several SMA actuator implementations in robotic systems have been reported. In this paper, we highlight selected examples that may serve as inspiration for future projects. Additional reviews of application possibilities can be found in other studies [24,80].

The structured comparison presented in Table 7 reveals that SMA-driven robotic systems can be broadly categorized into two practically distinct implementation approaches:

➢

compact, structurally simple mechanisms with minimal peripheral integration (commonly grippers and micro-manipulation systems), and

➢

performance-oriented, engineering-grade systems (primarily wearable devices and exosuits), where actuation is tightly integrated with sensing, control, and thermal management [116].

Table 7. Application of SMA actuators in robotics.

Robotic Subdomain	Application	SMA Type	Stroke, Deformation, Strain	Force, Torque	Reaction, Frequency	Cooling
Grippers	Monolithic compliant gripper for microassembly [101]	SMA wire (≈0.15 mm)	Max finger tip stroke 1.2 cm; Max SMA contraction 3%	SMA force 2.8 N, Gripper force 0.38 N	Not mentioned	Passive
Wearables/exosuits (rehabilitation)	Soft exosuit elbow with multi-bundle SMA [49]	multiple wires (bundle), wire ~0.51 mm	Rotation 0–120°, position error 0.03 mm	105 N	0.065 Hz, 28 s per cycle	Passive
Wearables/exoskeleton (SMA springs)	Soft bionic elbow exoskeleton [50]	SMA springs	Rotation 0–80°	48 N	0.13 Hz	Passive
Soft/biomimetic robots	Aurelia-Inspired Robot Based on SMA Artificial Muscles [52]	SMA “artificial muscle modules” from SMA wires	Not mentioned Locomotion	Not mentioned	2 Hz, max. velocity 12 cm/s	Water cooling
Robotic hands + microrobotics (SMA fibers/coils)	Functionalized SMA coil (PDA-AgNW) for robotic arm and microrobot [54]	SMA coil (SMAc), surface modified	Strain 40–200%	0.77 N (40 = strain), 1.72 N (200% strain)	1 Hz	Passive
Exoskeleton (ankle)	SMA-driven ankle exoskeleton [117]	SMA wire	Stroke 4 mm	150 N	0.5 Hz	Not mentioned

Table 7. Application of SMA actuators in robotics.

Robotic Subdomain	Application	SMA Type	Stroke, Deformation, Strain	Force, Torque	Reaction, Frequency	Cooling
Grippers	Monolithic compliant gripper for microassembly [101]	SMA wire (≈0.15 mm)	Max finger tip stroke 1.2 cm; Max SMA contraction 3%	SMA force 2.8 N, Gripper force 0.38 N	Not mentioned	Passive
Wearables/exosuits (rehabilitation)	Soft exosuit elbow with multi-bundle SMA [49]	multiple wires (bundle), wire ~0.51 mm	Rotation 0–120°, position error 0.03 mm	105 N	0.065 Hz, 28 s per cycle	Passive
Wearables/exoskeleton (SMA springs)	Soft bionic elbow exoskeleton [50]	SMA springs	Rotation 0–80°	48 N	0.13 Hz	Passive
Soft/biomimetic robots	Aurelia-Inspired Robot Based on SMA Artificial Muscles [52]	SMA “artificial muscle modules” from SMA wires	Not mentioned Locomotion	Not mentioned	2 Hz, max. velocity 12 cm/s	Water cooling
Robotic hands + microrobotics (SMA fibers/coils)	Functionalized SMA coil (PDA-AgNW) for robotic arm and microrobot [54]	SMA coil (SMAc), surface modified	Strain 40–200%	0.77 N (40 = strain), 1.72 N (200% strain)	1 Hz	Passive
Exoskeleton (ankle)	SMA-driven ankle exoskeleton [117]	SMA wire	Stroke 4 mm	150 N	0.5 Hz	Not mentioned

In the first category, publications typically emphasize geometric performance (e.g., tip displacement, jaw opening) and static force output, while dynamic control metrics are often only qualitatively illustrated. For instance, SMA-actuated compliant grippers report centimeter-scale jaw motion and sub-newton output forces, but response time, closed-loop bandwidth, and cycle frequency are not always reported in a standardized way, which limits direct comparison across designs [101].

In contrast, wearable robotic systems and exoskeletons tend to provide more explicit quantitative reporting, reflecting stricter safety and performance constraints. Multi-bundle SMA exosuits, for example, combine position and force control and explicitly address cooling-phase limitations via sequential activation strategies while also quantifying assistive performance and tracking behavior across repeated cycles [49,51]. Likewise, SMA spring-driven wearable architectures often disclose practical PWM implementation parameters (e.g., heating vs. holding duty cycles) and discuss cooling strategies as a primary determinant of achievable actuation frequency [50].

Soft and biomimetic robotic platforms provide an additional perspective on how environmental heat transfer conditions can relax thermal constraints. For example, underwater SMA-driven systems exploit strong convective cooling to support rhythmic locomotion under coordinated heating strategies (e.g., CPG-based control combined with adaptive regulation) [52]. Complementarily, recent “material–system co-design” approaches—such as functionalized SMA fibers/coils demonstrated in robotic hand and microrobot contexts—show that improving thermal/electrical interfacing can extend usable operational regimes while maintaining meaningful force output and repeatable cycling over short test horizons [54].

From a practical design perspective, the quantitative comparison underscores the importance of reporting standardized descriptors, including: SMA form (wire, spring, foil, coil), achievable strain or stroke, force or torque output, response time or cycle frequency and cooling method. Without such standardized reporting, meaningful comparison across robotic implementations remains difficult, and the “limitations” discussion (stroke limits, slow response) cannot be cleanly connected to real design choices [47].

Importantly, the data presented in Table 7 directly connect commonly discussed SMA constraints—limited recoverable strain for wire-based designs, thermally limited dynamics, and heat accumulation—to concrete robotic design decisions. Strategies such as multi-wire bundles, segmented activation, optimized PWM profiles, forced convection, liquid cooling, or active heat extraction schemes represent engineering attempts to overcome these intrinsic constraints and should therefore be discussed alongside the reported performance metrics [49].

In summary, while SMA actuators offer compelling advantages in compactness, silent operation, and high force-to-weight ratio, their practical deployment in robotics remains fundamentally constrained by thermally limited dynamics. A quantitative comparison across representative applications provides a clearer framework for assessing suitability and highlights the central role of integrated control and thermal management in advancing SMA-based robotic actuation [47].

3.1. Mobile Robots

An in-pipe mobile robot designed to move inside pipelines uses a helical spring–type SMA actuator as its propulsion mechanism. Locomotion is achieved through cyclic length changes in the SMA actuator, which drive the robot forward inside the pipe via bristles mounted diagonally on the robot body (Figure 23) [118,119].

An earthworm-inspired mobile robot (Figure 24) consists of silicone bellows that act as a bias spring to generate the restoring force. When the SMA actuator contracts, the silicone bellows deform and store elastic energy; upon deactivation of the SMA actuator, the stored energy drives re-extension of the actuator to its original length and shape. Front and rear needles create a difference in frictional forces between the robot body and the substrate, resulting in clean forward locomotion [120].

An earthworm-inspired robot (Figure 25) was constructed from multiple modules comprising a low-stiffness silicone shell, which simultaneously serves as a bias spring for a helical spring–type SMA actuator positioned at the center of each module. When an electric current is applied, the SMA actuator heats up, contracts, and deforms the silicone shell. After the current is removed, the silicone shell recovers its original shape and extends the SMA actuator back to its initial length [40,121].

An SMA actuator was also employed in a prototype eight-segment snake-like robot (Figure 26) [122]. The robot consists of a structure of rotary jointed segments, each driven by two antagonistically arranged SMA actuators to produce rotational motion. By applying an appropriate actuation sequence, net locomotion of the robot is achieved.

In the design of a snake-like robot (Figure 27), SMA actuators were used to drive the individual modules. Each module consists of a support board, a control board, and SMA wire actuators with bias springs to realize module contraction and expansion. Each module provides three degrees of freedom, enabling three-dimensional locomotion of the robot [123].

In the design of the wheeled robot (Figure 28), an SMA actuator is used, and its linear motion is converted into rotational motion of the wheels. The periodic linear displacement of the actuator produces a change in the axial distance between the wheels. Backward wheel rotation is prevented by a self-locking mechanism. This proposed concept of a resilient–rigid coupling SMA actuator (RRSA) represents an interesting approach to implementing SMA actuation [124].

A biologically inspired locomotion strategy—similar to an inchworm in nature (Figure 29)—is employed in the design of a soft crawling robot. An SMA spring actuator is used to generate periodic changes in the distance between the front and rear legs. The legs feature foot structures with directional friction, which ensures that the cyclic variation in leg spacing results in clean forward motion of the robot [125]. A similar principle is also employed in an untethered, mechanically intelligent inchworm-type robot driven by an SMA spring actuator [126].

3.2. Manipulators and Grippers

The robotic gripper (Figure 30), actuated by an SMA wire actuator, is designed as an end effector for grasping objects using an industrial manipulator. The overall configuration consists of three arms evenly distributed around the circumference. All arms are driven by a central linkage (pull rod) that is actuated by the SMA element. The arms are designed as a lever mechanism to amplify the displacement generated by the SMA actuator [127].

A similar approach is employed in another gripper design (Figure 31), which incorporates lever transmissions and direct actuation of a piston using SMA wire actuators. The system comprises a housing (1), three NiTiFe SE wires (2), a piston (3), three NiTiCu SMA wires (4), a fork (5), nine pins (6), six connecting rods (7), and three claws (8) that act as gripping elements. The NiTiCu wires (4) are in the martensitic state at room temperature, being pre-strained and connected with one end at the housing base and with the other end to the piston [128].

A two-finger gripper actuated by an SMA actuator (Figure 32) achieves an opening range of up to 40 mm, which is comparable to commercially available electric grippers. The SMA actuator is clamped in two holding blocks and routed through a multi-pulley system to reduce the required installation space. The inherently small stroke of the SMA actuator is converted into rotational motion via a lever arm, and the resulting motion is further amplified to the desired magnitude using a gear transmission. Finger opening is then realized through a parallelogram linkage. The design is optimized to minimize the overall installation dimensions of the gripper (Figure 32) [129].

A large-stroke SMA wire actuator (Figure 33) was developed for a robotic gripper. The mechanism consists of a set of eight sliding shafts driven by SMA wire actuators. The displacements of the individual sliding shafts are mechanically coupled so that the overall stroke is accumulated. Upon activation of the SMA actuators—which are electrically connected in series and heated by Joule heating via an applied current—the total output displacement is generated as the sum of the partial displacements of the individual SMA elements. The final sliding shaft serves as the mechanism’s output for producing the gripper’s grasping force. After deactivation and cooling of the SMA actuators, the last sliding shaft returns to its initial position by means of a bias spring. The gripper can handle objects with dimensions in the range of 60–100 mm. The mechanism incorporates eight NiTi SMA wire actuators with a diameter of 0.4 mm and a length of 100 mm [130].

A multi-bundle SMA actuator (Figure 34) [25,49] is a concept for obtaining a higher output force by arranging multiple SMA wire actuators in parallel compared with a single SMA wire. In a mechanically parallel configuration, each wire contributes to the total load-carrying capacity of the system. When the wires are arranged in parallel and uniformly tensioned, their tensile forces add up; thus, the total force is approximately the sum of the forces generated by the individual wires. Using multiple SMA wires increases the overall power/energy demand of the concept; however, this may not be a limiting factor in stationary applications. The concept has been employed in rehabilitation medical devices, but it is also applicable to robotic systems and other domains [131].

The elephant trunk robot (Figure 35) features an articulated structure inspired by the biological model of an elephant trunk, comprising serially arranged modules connected by rotational joints. A servomotor-driven cable is employed as the primary driving “muscle” of the robot, while an SMA actuator combined with a return spring is used as an auxiliary actuator to modulate the stiffness of the structure [132].

An SMA actuator was used in a prototype SMA-actuated flexible gripper with variable stiffness. The grasping action itself is provided by an electric motor, whereas stiffness adjustment is achieved using an SMA spring. By changing its temperature, the SMA spring alters its elastic modulus and, consequently, the overall stiffness of the gripper. The SMA spring is connected in series with the motor, thereby modulating the stiffness of the grasping mechanism. For this application, a trained SMA material exhibiting the two-way shape memory effect was selected, which does not require an external bias force to recover its original shape. The properties of this SMA spring are governed solely by temperature variation [133,134].

Shape memory alloy (SMA)-based soft actuators (Figure 36) have been developed as composite structures comprising a polymeric matrix, in which an SMA wire actuator is positioned eccentrically with respect to the neutral axis along the composite beam. This is a multilayer architecture incorporating different materials such as polymers, glass fiber, ABS, and others, thereby forming a bending-type SMA actuator [135].

A similar soft SMA actuator concept (Figure 37) has also been applied in the development of SMA-based artificial muscles. The actuator incorporates an SMA wire embedded within a silicone rubber layer together with a hydrogel and thermoelectric materials. Heating is achieved via Joule (resistive) heating. The hydrogel serves as a passive cooling medium, while the integrated thermoelectric material actively removes heat from the SMA wire. The authors used this concept to develop a three-finger gripper [136].

Another composite structure (Figure 38) is composed of a two-way shape memory alloy (SMA), a low-temperature epoxy, glass fiber–reinforced plastic (GFRP), and fiber-reinforced plastic (FRP). This structure is used in a prototype gripper for grasping objects of various shapes [137]. Other similar solutions have also been developed and published in the papers [110,138,139,140,141,142,143,144].

3.3. Biomedical Application

Titanium alloys are widely used in medicine as prosthetic materials due to their excellent strength and biocompatibility. The solutions developed in this domain can also be applied in robotics; therefore, they are discussed here.

Shape memory alloys are used in the design of a hand prosthesis. Wire actuators are used, and forced air cooling was proposed to improve actuator dynamics. A PID controller was implemented for position control. The resulting performance of the prosthetic hand is comparable to that of other commercially available devices [103,145]. Additional similar solutions have also been developed and reported in the papers [146,147,148,149].

To address hand paralysis, wearable exo-gloves (Figure 39) have been developed to enable telerehabilitation and training during the treatment of patients with such impairments. In this case as well, shape memory alloy actuators were used to control the motion of the exo-glove’s five fingers [116].

Stroke is a major cause of health complications, including motor impairments and other deficits that can prevent individuals from living independently. Conventional exoskeletons exist for rehabilitation purposes; however, they have several limitations. To support affected individuals, research and development has focused on a new soft exoskeleton intended for elbow rehabilitation, in which motion and force are generated using a shape memory alloy actuator in the form of a multi-wire bundle [49].

Stapes prostheses use smart actuators based on shape memory alloys. These prostheses are used to restore hearing in patients with otosclerosis. The material provides excellent sound transmission, offering significant potential for such applications. Future development is expected to be closely linked to additive manufacturing (3D printing) technologies. Research in micro-electro-mechanical systems (MEMS) will provide substantial opportunities for implantable middle-ear transducers with integrated sensors and actuators. Embedded systems implemented within MEMS devices will enable advanced acoustic signal processing, such as limiting exposure to extreme noise levels, among other functionalities [150,151].

The authors also used an SMA wire actuator in the design of a finger prosthesis, where it functions as a flexor “muscle” to bend the finger during heating of the SMA element. A superelastic wire is used to provide a bias force that returns the finger to its initial position [152].

Solutions involving SMA actuators have also been developed for minimally invasive surgical procedures [75,153,154,155,156].

3.4. Industrial Approaches to the Implementation of SMA Actuators

In recent years, the field of shape memory alloy (SMA) actuators has shown a gradual shift from purely laboratory prototypes toward application-oriented and partially standardized solutions. This trend is particularly evident in studies focusing on modularity, repeatable manufacturing, and the integration of SMA actuators into complex mechatronic and robotic systems.

From a design-principle perspective, industrially oriented SMA actuators can be classified into several fundamental categories. The most widespread approach is based on linear SMA actuators using wire or multi-wire elements that generate tensile force during the martensite-to-austenite phase transformation. This concept is generally regarded as the most technologically mature, as it enables a high force-to-mass ratio with relatively straightforward mechanical integration and is compatible with conventional design practices in mechanical engineering and robotics [10,12,87,157].

Another important direction involves planar and foil-based SMA actuators that utilize thin SMA layers or strips integrated directly into load-bearing structures. These solutions are particularly suitable for miniaturized systems, where low mass, fast thermal response, and actuator integration without additional transmission mechanisms are critical. In application-driven literature, this approach is often associated with microrobotics, precision positioning, and compact robotic mechanisms [158,159].

A distinct category comprises SMA actuators designed as part of a controlled mechatronic module, in which the physical behavior of the SMA material is tightly coupled with the modeling and control algorithms. These solutions emphasize repeatability, stability, and accurate positioning, often using sensorless control, state-variable estimation, or hysteresis compensation. Such an approach is typical of applications targeting the practical deployment of SMA actuators in robotics and industrial equipment [47,91,160].

From an industrial perspective, an increasing number of studies also focus on functionally defined SMA actuator modules. In these cases, the SMA actuator is designed as a ready-to-use functional unit with a clearly specified mechanical output, thereby lowering the barrier to implementation in practical applications. Although such solutions do not yet reach the level of standardization typical of electric or pneumatic actuators [10,63].

Overall, current industrial approaches to implementing SMA actuators are trending toward modularity, application-specific specialization, and the integration of control directly into the actuation system. This development indicates that SMA actuators are gradually transitioning from experimental demonstrations to practically deployable drive elements for specific robotic and mechatronic applications.

4. Discussion

4.1. Future Development of SMA Actuator Applications in Robotics Research Areas

A temporal trend analysis of publications indexed in the Web of Science (WoS), Scopus, and IEEE Xplore databases indicates a long-term and steady increase in research on SMA actuators in robotics. Whereas studies in the 1980s and 1990s were predominantly isolated and conceptual, a pronounced rise in publication activity has been observed after 2010. The strongest growth occurs after 2020, with the period 2022–2025 exhibiting the highest annual publication output. This trend suggests that SMA actuators remain a promising and actively developing technology in robotics. It is closely linked to significant progress in materials research, including the development of alloys with lower transformation temperatures.

The most significant growth in research interest has been observed in the following SMA actuator areas:

• Additive manufacturing and advanced fabrication technologies, including 3D printing, laser processing, lattice structures, and multi-material integration of SMA actuators.
• Soft robotics and compliant structures, where SMA actuators are used to achieve muscle-like behavior and safe interaction with the environment.
• Microrobotics and micromachines, often without explicitly using the term “robot”, yet with clearly identifiable mobility or functionality within robotic systems.
• Closed-loop control and sensorless control, aimed at compensating hysteresis, nonlinearities, and the temperature dependence of SMA actuators.
• Composite and multifunctional material solutions, combining SMA elements with polymers, fibers, or multilayer structures.

Based on an analysis of the most recent publications (2022–2025), future research and development of SMA actuator applications in robotics can be structured into the following main thematic areas:

• Soft grippers and compliant robotic hands: Research focuses on adaptive grasping, bioinspired manipulation, and safe human–robot interaction using SMA-driven fingers and end effectors.
• Microrobots and micromachines: Applications are dominated by miniature mobile systems, untethered microrobots, and microdevices, often extending to medical and inspection tasks.
• Advanced manufacturing of SMA robotic systems: Increasing emphasis is placed on additive manufacturing, laser-based technologies, and the integration of SMA elements into modular robotic structures suitable for repeatable production.
• Compact joints and actuator modules: SMA actuators are increasingly developed as integrated bending or rotary modules for lightweight manipulators, continuum robots, and modular robotic architectures.
• Wearable and rehabilitation devices: A substantial portion of work targets rehabilitation gloves, exoskeletons, and assistive devices, where quiet operation and a high power-to-mass ratio are key advantages.
• Control and sensing in robotic systems with SMA actuators: Although fewer in number, these studies address a technologically critical area focused on precise position control, state estimation, and improved motion repeatability.
• Terrain and inspection robotics (in-pipe, climbing, and crawling robots): These applications remain relevant in specific scenarios where compactness, compliance, and the ability to operate in confined spaces are decisive.

SMA actuators are highly attractive due to a range of favorable properties; however, a major challenge remains the position control of SMA actuators. Multiple control approaches have been proposed, and development in this area is ongoing. In [68,132,133,134], a deep reinforcement learning technique was applied to control the temperature and stroke of an SMA actuator used in a gripper. Heating of the SMA actuator was achieved via Joule (resistive) heating, while cooling was provided by cooling fans. The use of artificial intelligence techniques appears to be one of the promising options for SMA actuation control, and future research should further explore these control strategies.

Significant progress in SMA actuator development is also enabling the emergence of wearable applications that support people in everyday life. Wearable systems with smart actuators often improve user comfort and, in some cases, address medical needs and rehabilitation. Current developments include wearable thermoregulatory textiles, wearable robots, medical-care devices, orthopedic aids, and haptic interfaces. The integration of smart actuators into textile structures represents a substantial opportunity for future research and innovation [161,162,163,164].

4.2. Industrial Perspectives on SMA Actuators in Robotic Applications

The analysis presented in this study indicates a clear convergence between industrial approaches to implementing shape memory alloy (SMA) actuators and current development trends in robotic applications. Although SMA actuators have long been perceived primarily as a laboratory or experimental technology, recent literature suggests a gradual shift toward application-oriented, modular, and integration-driven solutions that are compatible with the requirements of industrial deployment.

From an architectural perspective, the most technologically mature industrial solutions are linear SMA actuators based on wire or multi-wire elements. These design principles also dominate robotic applications such as grippers, compliant robotic hands, and bioinspired manipulation systems. This overlap suggests that key requirements for industrial feasibility—ease of integration, compactness, and a high force-to-mass ratio—are consistent with robotics requirements for safe interaction and compliant behavior.

Planar and foil-based SMA actuators, which in industrial contexts are primarily associated with miniaturization and direct integration of actuation into load-bearing structures, show a strong relation to microrobotics and compact robotic mechanisms. In such systems, the boundary between the actuator and the structural element becomes less distinct, which aligns with robotic trends toward multifunctional and highly integrated components.

A further major point of convergence is represented by industrial solutions that emphasize integration of control directly into the SMA actuator. The growing number of robotic applications employing sensorless control, hysteresis compensation, and model-based regulation indicates that control is becoming an inseparable part of the actuator itself. This approach is essential to achieve repeatability, stability, and motion accuracy, which are prerequisites for practical robotic and industrial use.

The trend toward functionally defined SMA actuator modules can be viewed as a direct response to the needs of robotics, particularly in rapid prototyping, service robotics, and wearable systems. Modularity and abstraction of the complex material behavior of SMAs enable their use in applications where detailed handling of actuator physics is undesirable, thereby lowering the barrier to wider acceptance.

Despite the identified progress, SMA actuators do not yet represent a universal replacement for conventional electromechanical or pneumatic drives. Limitations related to energy efficiency, response speed, thermal management, and fatigue life under cyclic loading continue to constrain broader deployment. Nevertheless, the current coexistence of industrially oriented solutions and rapidly expanding robotic applications suggests that SMA actuators have reached a level of technological readiness suitable for specialized, high-value use cases.

Overall, the future impact of SMA actuators in robotics will depend on the continued co-evolution of materials research, actuator design, control strategies, and system-level design of robotic applications. The industrial perspectives discussed here provide important context for understanding how SMA actuators may gradually transition from experimental concepts to reliable functional elements in the next generation of robotic systems.

5. Conclusions

This review study provided a systematic overview of the development and current state of shape memory alloy (SMA) actuator applications in robotics, based on an extensive publication analysis of the Web of Science, Scopus, and IEEE Xplore databases. The combination of deduplication, thematic classification, and content analysis of abstracts enabled the identification of not only the dominant application domains but also long-term trends and emerging research directions.

The results clearly show that SMA actuators maintain strong and stable research relevance in robotics, with publication activity increasing markedly in recent years. While earlier studies were primarily focused on demonstrating fundamental actuation principles, current research is shifting toward the integration of SMA actuators into functional robotic systems that emphasize compliance, compactness, and bioinspired behavior.

The most significant application areas include soft robotics, robotic grippers and hands, microrobotics, and wearable robotic systems, where SMA properties—high power-to-mass ratio, silent operation, and the ability to generate complex motion within simple mechanical structures—provide significant advantages over conventional actuation technologies. At the same time, the importance of advanced manufacturing technologies, such as additive manufacturing and multi-material integration, is increasing, enabling SMA actuators to be realized as compact modules suitable for repeatable production.

Content analysis of papers indicates a growing emphasis on closed-loop control, sensorless control, and hysteresis compensation, reflecting efforts to overcome the principal limitations of SMA actuators associated with nonlinearity, slower dynamics, and thermal dependence. These control and modeling approaches are increasingly becoming integral components of robotic systems using SMA actuators, thereby improving accuracy, repeatability, and practical usability.

Despite the progress identified, several open challenges remain. These include improving energy efficiency, accelerating cooling, extending lifetime under cyclic loading, and standardizing actuator modules for broader deployment in robotic platforms. Future research should focus on the holistic design of robotic systems, in which SMA actuators, the mechanical structure, control, and manufacturing technology are conceived as an interconnected whole.

In conclusion, SMA actuators do not constitute a universal replacement for conventional actuators; however, in specific robotic applications—particularly in soft, miniature, and wearable robotics—they offer unique properties that make them a key technology for the development of the next generation of adaptive and bioinspired robotic systems.

From a technology-readiness perspective, it should also be noted that SMA actuators are no longer exclusively confined to academic research. In recent years, the first commercially available SMA actuator modules have emerged, offered as off-the-shelf components with defined mechanical and electrical specifications. Although these industrial solutions do not yet reach the level of standardization typical of electric linear actuators or pneumatic cylinders, they demonstrate the increasing technological maturity of SMA actuators and their readiness for deployment in specialized applications. In particular, in domains where compactness, silent operation, low mass, and direct integration of actuation into the mechanical structure are decisive, SMA modules represent a viable alternative to conventional actuation technologies. This trend suggests that further development of SMA actuators will likely progress not only toward new robotic applications but also toward greater modularity, standardization, and industrial use.