Soft Optical Waveguides for Biomedical Applications, Wearable Devices, and Soft Robotics: A Review

In the domains of biomedical applications, wearable devices, and soft robotics, recent advancements have underscored the potential of soft, stretchable, and biocompatible devices. The design of optical soft devices has emerged as an ideal candidate for many applications owing to their high flexibility and immunity to electromagnetic interference. In this review, recent advances in soft optical waveguides, including advanced material selection, fabrication strategies, and characterization, are discussed. Herein, a comprehensive summary of the soft‐waveguide sensing strategies and actuation approaches are provided. Furthermore, the extensive applications of soft optical waveguides in the fields of biomedicine, wearable devices, and soft robotics are explored. Lastly, the challenges and opportunities for the future of soft optical waveguides, including multimodal sensing, algorithm optimization, and manufacturing scalability, are discussed.


Introduction
11][12][13] Typically, sensors and actuators are implemented using electronic strategies that depend on the relative variation of their electrical properties (e.g., piezoelectric, capacitive, or resistive).These sensors have good sensitivity and offer a simple readout. [14]However, electrical devices usually have poor biocompatibility and are sensitive to electromagnetic interference. [15][18][19] Soft optical waveguides not only exhibit a remarkable resistance to electromagnetic interference but also offer exceptional flexibility and durability.In particular, compared with chemically non-biocompatible soft materials, soft optical waveguides utilize biocompatible materials, such as hydrogels, naturally derived polymers and elastomers, [20][21][22][23][24][25][26] making them particularly suitable for long-term wear or implantable applications within the human body.Additionally, soft waveguides use light as a carrier of information and energy, enabling highly sensitive sensing and actuation in multiple dimensions (mechanical, thermal, biochemical, etc.).
In this review, we discuss important developments and advancements in sensors and actuators based on soft optical waveguides, which have laid a robust foundation for the application of soft opticalwaveguide technology in the fields of biomedical applications, wearable devices, and soft robotics.The first component provides a broad overview of different materials, structures, and preparation strategies for the various soft optical waveguides currently under investigation.The subsequent sensing component reviews soft optical-waveguide sensors based on different optical parameters (such as light intensity, pattern, wavelength, and phase).The review then discusses soft-waveguide-based actuators, which usually work based on photochemical reactions or photothermal effects.We also highlight their potential for diverse applications in biomedicine, wearable devices, and soft robotics.In the last part, we discuss the challenges of soft waveguides, pointing out their potential future directions in the scientific community.

Soft Waveguide Materials: Fabrication and Characterization
[29][30] However, they often exhibit stiffness and brittleness, non-biocompatible, which can cause an immune response, potentially resulting in tissue infections when employed within the human body.Moreover, conventional optical-waveguide sensors are non-biodegradable, requiring removal via secondary surgical procedures if left in the body for prolonged periods.These constraints render them unsuitable for applications involving wearable and implantable sensors, where flexibility, biocompatibility, and sustained long-term functionality stand as paramount considerations.As an alternative, there has been an increased quest to develop soft waveguides that can simultaneously accommodate high optical transmission and biocompatibility.

Materials
A variety of materials can be utilized to construct soft optical waveguides, including hydrogels, naturally derived polymers, and elastomers.A hydrogel is a gel-like structure that is highly hydrophilic, with a 3D cross-linked network.It can rapidly swell in water while maintaining its structural integrity.[40][41][42][43] These materials are biodegradable and can be decomposed and absorbed by the body.Elastomers, which are natural or synthetic polymers with elastic properties, can deform elastically under tensile and compressive stresses, and then return to their original shape. [44,45]These materials are highly flexible and transparent, making them suitable for use not only as complete optical waveguides, but also as cladding or encapsulation for soft optical wearable devices.
In addition, compared to conventional solid optical waveguides, the core of liquid-based waveguides offers higher level of flexibility than solid.The optical characteristics of these waveguides can be precisely customized by controlling the optical properties of the liquid core materials.The combination of a liquid core and polymer cladding opens new avenues for exploring the capabilities of soft optical waveguides.Techniques such as soft lithography [46] and photocurable liquid core-fugitive shell printing [47] have been employed for the fabrication of these innovative optical waveguides.Table 1 summarizes some representative soft optical materials and strategies for waveguide fabrication.
Apart from their favorable biocompatibility and elasticity, the refractive index and optical loss are critical properties for soft waveguides. [48]The optical loss is primarily attributed to the inherent optical loss of the material and the optical structural loss caused by the refractive index distribution of the core and cladding.On the one hand, the light absorption of materials and nonuniform of waveguides lead to the inherent optical loss of soft optical waveguides.[51][52] Total internal reflection is imperative for highly efficient optical waveguides. [53]The refractive index of the core material must surpass that of the cladding.The cladding can consist of a low refractive index coating or an external environment with a low refractive index, such as air, water, or biological tissue.

Fabrication and Characterizations
Soft waveguides come in two main structures: large-sized and microstructured planar optical waveguides.Microstructured planar soft waveguides typically have complex and diverse waveguide structures that are tens of microns in size.In recent years, inkjet printing, [54][55][56] laser direct writing, [57,58] and photolithography [59,60] have emerged as efficient techniques for fabricating these structures.Common microstructures found in optical waveguides are diffraction gratings.When external stress is applied, the structural dimensions of this flexible grating are slightly altered, which in turn affects its optical transmission characteristics, especially the reflection wavelength.As shown in Figure 1a, the fabrication of 3D diffraction gratings using hydrogels that were structured by thermal nanoimprint lithography has been demonstrated. [59]These hydrogels are sensitive to pH and water, making them ideal sensors for stimuli-responsive applications as optical grating change their morphology, ultimately impacting their optical properties.
To effectively confine the light, the cladding can be further fabricated using dip coating. [61,64,65]As shown in Figure 1b, PDMS (polydimethylsiloxane) precursors with distinct curing agent ratios, leading to different refractive indexes (RI), were employed for the core and cladding.The fiber core was created by injecting the high-RI PDMS precursor (base/curing agent, 5:1) into a silicone tube mold and subsequently undergoing thermal curing.Following the removal of the core from the mold, it was immersed in a low-RI PDMS precursor (base/curing agent, 20:1) to apply a viscous cladding layer.The utilization of 3D-printing technology has demonstrated its capability in creating molds of complex geometries for millimeter-scale optical waveguides.This method has proven to be effective in the fabrication of large-scale optical waveguides with varied shapes.For example, Zhao et al. used 3D-printed molds to fabricate a strain-sensing optical waveguide, [66] as shown in Figure 1c.The core material was a transparent polyurethane rubber with a refractive index of 1.461, while the silicone composite cladding had a refractive index of 1.389 and exhibited a high propagation loss of 1500 dB cm À1 at 860 nm.The silicone composite was selected as cladding for its high absorptivity and low refractive index, promoting total reflection of light and preventing interference from ambient light.This optical waveguide was employed for sensing bending, stretching, and compression and was integrated into a fiber-reinforced soft prosthesis for a series of active haptic experiments.
In addition to the previous methods, soft optical waveguides with regular structure can be easily formed by the extrusion of long filaments, making this method particularly suitable for manufacturing long soft waveguides. [24,47,61,67]Encouraging results have been achieved through the combination of filaments extrusion and 3D-printing technology for fabricating soft optical waveguides. [68]Heiden et al. 3D printed gelatin-based hydrogel (biogel) inks into dimensionally stable complex objects to form stretchable optical waveguides, as shown in Figure 1d.The printed optical waveguides can be extended to five times their original size.The printed biogel are reusable and completely biodegradable, which greatly improves the utilization of the material and reduces the environmental problems.
In summary, inkjet printing, laser direct writing, and photolithography techniques excel in the fabrication of microstructured waveguides, offering precision and control at the microscale.Methods such as molding, extrusion of long filaments, and flow casting are better suited for the manufacturing of large-sized optical waveguides, allowing for more efficient and cost-effective large-scale production.Nevertheless, it is important to note that these techniques may come with the disadvantage of limited resolution, which can impact the fine details and intricacies achievable in the waveguide structure.
these variables can be used as sensing variables to sense information about the external environment.

Optical Intensity
Soft waveguides have different loss mechanisms due to the complexity of light transmission.If the output power of a soft waveguide with no bending, stretching, pressing, and any other deformation is defined as the baseline intensity I 0 , the output power loss is defined as where I is the measured intensity.By this definition, the output intensity loss I loss is 0 with no deformation.As the deformation increases, the output loss becomes greater than zero.[71][72][73][74][75][76] Han et al. fabricated a tactile sensor based on a silicone material (Dragon Skin 10) combined with two fibers. [77] clearance δ at the elongating edge is designed to measure deformation.Larger lateral displacements cause larger gap, increasing δ and causing a decrease in transmitted light intensity.Figure 2a shows the intensity variation under dynamic stick-slip situation.The pebbles hit the bottom of the bottle, making a transient impulse signal.Without the need for force measurements or external image recognition, the purposed sensor can measure friction information and grasp objects according to the intensity changes.Levi et al. embedded eight infrared emitters and eight photodetectors into a 5.5 cm diameter planar PDMS waveguide. [78]When the material was touched, the optical signals experience loss due to the material deformation and frustrated internal reflection.The output signal was converted to an electrical signal and the pressure map was reconstructed which can display not only the contact position but also the contact intensity (Figure 2b).Dong et al. also proposed a graphene PDMS sensor which can detect different levels of strain including limb bending and muscle relaxation. [79][82][83][84][85] Heiden et al. proposed a 3Dprinting process based on fused deposition modeling to fabricate stretchable waveguides capable of both proprioceptive and extrinsic sensations. [68]The proposed soft waveguide is a bare waveguide without cladding.Hydrogels allow refractive index adjustment through water content modification.This device was capable of dynamic real-time control and can automatically seek and wipe to detect and remove obstacles.Based on light scattering and absorption from three chambers, the output light intensity signal can respond to different deformations (touch and bending).Figure 2c displays the change in light intensity for different deformations.The response times are within 0.63 s at 50 kPa applied pressure.However, it can be challenging to distinguish when multiple events occur simultaneously.Reproduced with permission. [77]Copyright 2022, Wiley.b) Design layout of the polydimethylsiloxane (PDMS) sensor and the position of four subsequent contacts.Reproduced with permission. [78]Copyright 2013, MDPI.c) Performance of multidirectional self-sensing actuators.Reproduced with permission. [68]Copyright 2022, American Association for the Advancement of Science (AAAS).d) Two-chamber-waveguide signal strength during single and combined strain modes.Reproduced with permission. [80]Copyright 2022, IEEE.e) Performance of soft-waveguide sensor for autonomously selfhealing elongation and flexion sensing.Reproduced with permission. [86]Copyright 2022, AAAS.

Krauss et al. guided a single light source into two connected
chambers. [80]Through the output power loss and magnitude of two chambers, the sensor can simultaneously sense the amplitude and direction of twisting and bending, as well as elongation, pressing force, and location.As shown in Figure 2d, the output of the composite deformation corresponds qualitatively to the summation of the bending and twisting.At the same time, some reasons for manufacturing inaccuracy are listed in the article, including material mixing ratio, position of light source and detector, tiny air bubbles from casting, and smoothness of the waveguide surface.
In addition, there are extreme modes of sensing for applications where device longevity is preferred, such as wearables for human-machine interactions.Bai et al. proposed an autonomous self-healing polyurethane urea elastomer based on the output light intensity which has inherent resistance to damage. [86]The article showed the change of light intensity from the beginning of the cut to the full recovery.When multiple perforation tests are performed on the sensor, the sensor becomes more sensitive in bending and strain tests instead because voids render them more lossy waveguides.To eliminate sensor drift and hysteresis, they designed the sensor in a wave shape to adapt amplitude and period of the strain.As shown in Figure 2e, the sensor has reliable dynamic sensing in 140 cycles of continuous cyclic testing.They also make a soft quadruped which can sense damage and heal by themselves within minutes.Leber et al. demonstrated a thermoplastic elastomer that can reversibly maintain 300% strain. [83][89] Chen et al. pre-injected photothermal nanoparticles to induce photothermal heating around the tumor. [67]When the temperature of the tumor site was close to the critical solution temperature of the sensor, phase separation occurred between the polymer network and water molecules within the core hydrogel fibers, the transparency decreased hinder light transmission through the sensor.
The photoconductivity can be recovered after cooling.Larger wavelengths would have lower attenuation, and larger diameters result in reduced light loss.Figure 3a showed good cyclicity about the thermally induced light-switching process.These results are promising to be broadly applied in the field of intelligent photomedicine.It also makes it possible for wearable devices to measure parameters such as the number of calories burned during exercise.Yetisen et al. covalently incorporated 3-APBA molecules into the core for glucose sensing. [61]The absorption of light by glucose was a minor contributor to the intensity changes.Upon boronic acid-cis-diol glucose complexation, the Donnan osmotic pressure increased and decreased the difference of refractive index between the core and cladding (Figure 3b).Hence, the intensity of transmitted light across the hydrogel sensor increased.The propagation loss across the fiber decreased 8.3% in 45 min under 100 mmol L À1 glucose.The change in transmitted light intensity was related to the concentration of glucose, showing a sensitivity of 1.2 mmol L À1 .Hydrogel soft waveguide shows potential to provide label-free sensing for continuous in vivo glucose monitoring for diabetic patients in clinical and care settings.
Overall, the light intensity sensing method is simple and easy to implement.However, since changes in ambient light can interfere with sensor measurements, the light source and environment requirements are stringent, and the unexpected deformation of the soft waveguide will cause errors.

Wavelength
Sensors utilizing wavelength as the sensing variable necessitate wavelength sensitivity to fluctuations in the external environment.[92] This approach allows for the creation of highly targeted, flexible sensors using different materials.Guo et al. made a conical hydrogel waveguide by doping carbon dots (CDs). [90]As shown in Figure 4a, this  [67] Copyright 2022, Springer Nature.b) Performance of glucose-sensitive hydrogel optical fibers under different time and concentrations.Reproduced with permission. [61]Copyright 2017, Wiley-VCH.
CDs-doped hydrogel waveguide exhibited high selectivity for Hg 2þ ions.It showed decreased fluorescence intensity with increased Hg 2þ concentration from measurements of the emission spectra.They also fabricated a similar hydrogel waveguide by doping glutathione-capped quantum dots (emitting at 615 nm), and their ability to exhibit high selectivity and sensitivity to Pb 2þ ions is demonstrated in Figure 4b. [91]Additionally, they introduced another type of quantum dots capped with thioglycolic acid (emitting at 528 nm), which is insensitive to metal ions, into the optical waveguide.The detection of Pb 2þ ions was quantitatively achieved through the measurement of the ratio of optical intensity between the two quantum dots.
Bragg gratings are a common waveguide structure, that respond to characteristic wavelengths.The fabrication process for Bragg gratings within traditional glass fibers is well established.The most efficient method for producing a flexible grating sensor is achieved by directly embedding a fiber Bragg grating (FBG) into a flexible material. [93,94]A flexible sensor, as depicted in Figure 4c, was designed by embedding four FBGs into an optical fiber within soft silicone material. [95]The silicone acts as a mediator, transferring the applied load to the embedded FBGs and causing a slight deformation within the FBGs.This deformation results in a shift in the reflected wavelength of the grating, as displayed in Figure 4d, which is dependent on both the applied load and the indentation's position.
However, although embedding optical fibers has substantially improved the rigidity of conventional options, the risk of fiber fragmentation persists.Alternatively, direct fabrication of Bragg gratings within flexible materials is a more feasible solution.An artificial skin with a grating structure made from a fully stretchable PDMS material has been recently fabricated. [60]This optical artificial skin has a tailored sensitivity that can detect compression forces ranging from 0 to 3.8 N, while also allowing for monitoring of deformations with up to 135% elongation.Particularly, a special mechanism for detecting the strain applied to the waveguide is the direct colorimetric method, which is essentially a wavelength measurement.Using a multicolor light source as the input light, the reflected wavelength change caused by the grating deformation can be observed by the change in the color of the output grating.

Other Sensing Strategies
Time of flight (TOF) is a common wave-based spatial sensing strategy used to measure the time delay of signal transmission and reception.The time-domain measurement of TOF optical  [90] Copyright 2017, Springer Nature.b) Specific wavelength sensing of Pb 2þ ions based on quantum-dots-doped tapered hydrogel waveguide.Reproduced with permission. [91]Copyright 2018, Springer Nature.c) Contact location and force sensing based on four fiber Bragg gratings embedded in soft material.Reproduced with permission. [95]Copyright 2020, Mary Ann Liebert.d) Wavelength and color sensing of compression and deformation based on PDMS waveguide with grating structure using e-beam lithography.Reproduced with permission. [60]Copyright 2022, MDPI.e) Phase shift sensing of temperature based on gel polymer waveguide.Reproduced with permission. [99]Copyright 2021, American Chemical Society.
sensing can be used for spatial sensing, but frequency-domain reflectometers are more common. [96]TOF can also be applied to measure optical reflection from semitransparent biological tissues, and optical coherence tomography, [96] and reconstruct 3D images at the microscale.When a soft optical waveguide is stretched or compressed, the TOF of the light transmitted in the waveguide changes accordingly.Therefore, the microbending of the waveguide induced by pressure not only leads to optical loss but also changes the TOF. [97,98]Lin et al. compared the response of flexible optical waveguides (polyurethane core and silicone resin cladding) to bending, pressure, and tension based on intensity and TOF sensing strategies. [97]The intensitybased sensing strategy is more suitable for pure bending scenarios, while the TOF-based sensing strategy can distinguish between pressure and tension and excels notably in distance measurement.
Furthermore, phase can serve as an additional sensing variable to wavelength sensing.Wang et al. developed and produced a multimode response optical waveguide sensor that employs a dependable, cross-linked gel polymer electrolyte. [99]The phase modulation of the optical signal is triggered by changes in ambient temperature, and is measurable by a photodetector, as shown in Figure 4e.The sensor is highly sensitive to temperature and humidity, boasting a sensitivity of 0.5π rad °CÀ1 within the monitored temperature range of 36.0-38.0°C.
In addition, integrating multiple sensing strategies in soft optical waveguides has more potential in complex application scenarios.For example, compared to single-variable sensing, combination of wavelength (color) and intensity sensing is capable of distinguishing complex mechanical deformation. [72,100]hese advancements in sensing fusion provide more comprehensive and accurate physiological information for health monitoring and disease diagnosis.

Soft-Waveguide Algorithm
Before machine learning (ML), people analyzed time series and signals by creating mathematical models, for example, analyzing the light in the time domain and frequency domain. [99,101]ultimodal sensing generally relies on different variables; a simple mathematical model cannot easily implement multimodal sensing.Artificial intelligence in data analysis, especially ML, can compensate for the limitations of individual sensors, discover hidden features in the data, and improve overall measurement accuracy. [102]For example, Massria et al. detected the location and intensity of an applied load on soft tactile sensor using the combination of numerical finite element method (FEM) and ML. [95]As shown in Figure 5a, they used FEM model data to train a cascade of two feedforward neural networks.Based  [95] Copyright 2020, Mary Ann Liebert.b) Design and sensing result of the proposed multifunctional soft sensor.Reproduced with permission. [84]Copyright 2020, American Association for the Advancement of Science.c) Conceptual schematic and simultaneous realization of sensing and perception.Reproduced with permission. [103]Copyright 2022, Springer Nature.
on the generalization ability of the neural network, the model accurately predicted the tactile position and force.Kim et al. generated three heterogeneous sensing mechanisms in one structure and used threshold evaluation and simple neural network to classify eight combined deformation modes. [84]The deformation consists of three independent modes (stretching, bending, and compression).As shown in Figure 5b, ionic liquid channel is only sensitive to stretching and compression, as the resistance is only sensitive to changes in channel length or cross-section area.The conductive fabric layer only responds stretching significantly.The optical signal is used to estimate the magnitude of each deformation mode because it is more sensitive than the other two signals.The accuracy for different deformation mode classification is higher than 95%.This provides the possibility to design and control interfaces for complex operational tasks in complex environments.
When using complex images such as speckles for sensing, deep learning algorithms are required to accurately predict the sensing information.Shimadera et al. relied on the utilization of a two-layer convolutional neural network (CNN) to encode and decode speckle patterns. [103]The speckle refers to complex interference patterns that are created by light scattering.CNN is a class of neural networks that specializes in processing patterns.The proposed method output changes are mainly due to the absorption and scattering of the silicone gel.The encoded patterns can then be decoded to extract information about the surface properties.The down sampling kernel size was adjusted to be close to the mean speckle size to minimize the impact of environmental factors such as vibration and air fluctuation on speckle patterns.The silicone material is deformed using a stainless steel cylindrical.Figure 5c shows that the estimated error using the proposed model is smaller than the estimated error of the regression model.Based on this method, it is possible to do multimodal sensing including shear force, distortion, and 3D shape of a contact object.Although the proposed approach can avoid complex integration of multiple sensing elements, the number of the training data is low.In addition, more complicated modes of deformation can be studied using deep learning in the future.

Soft-Waveguide Actuator
Effective self-driven actuators are essential for developing functional bio-robotic systems.[106][107][108] Photothermal actuators convert photosensitive material absorbed optical energy into thermal energy.Thermal energy produces mechanical displacement of a structure based on the properties of the material (coefficient of thermal expansion [CTE], molecular adsorption, etc.).Xiao et al. demonstrated a repeatable photoactuator by embedding optical fiber taper into a PDMS/Au nanorod-graphene oxide photothermal film. [109]nlike free-space light-driven actuators, the light supply was located inside the material and controlled in real time (Figure 6a).When a 635 nm light was introduced to the elastomer, due to the high CTE of PDMS and low CTE of graphene oxide layer, this mismatch leads to a more than 270°bending of the photoactuator.Based on the characteristics of large bending angle and fast response, this photoactuator can grasp objects of different sizes, shape, and weight.
For example, light-responsive hydrogel waveguide actuators could be achieved by adding photoreactive materials to polymer networks.When exposed to light, photoreactive materials undergo photoreactions (such as isomerization, dimerization, cleavage, and rearrangement, etc.) to convert captured light signals into chemical signals that are channeled into polymer networks.This will change the properties of the hydrogel and thus actuate the hydrogel waveguide.Azobenzene is a representative photoreactive functional group, which will be isomerized under light irradiation. [113]As shown in Figure 6b, when it is exposed to ultraviolet light, the azobenzene will undergo isomerization from trans-to cis-form, which makes the hydrogel waveguide expand.When it is exposed to visible light, azobenzene will undergo reverse isomerization from cis-to trans-form shrink the hydrogel waveguide. [114,115]ith the special design of actuator architecture and implementation of advanced materials, we believe that these actuators can be deployed in object manipulation, microrobotics, biomedicine, and other applications.In scenarios such as in vivo controlled drug release, noninvasive surgical treatment of stroke, and underwater robotic control, soft-optical-waveguide actuators have significant potential advantages due to its contactless, fast response time, and high accuracy.

Applications 6.1. Biomedical Detection and Therapeutics
Soft optical waveguides, characterized by their small size, high precision, and minimal side effects, are extensively utilized in the field of medical diagnostics and therapy. [116,117]ocompatible soft optical waveguides, when implanted in the body, do not trigger cytotoxic reactions or immune responses, rendering them highly promising for applications in implantable medical devices.The soft optical waveguide exhibits the ability to traverse body tissues without causing harm to adjacent organs.120][121] Implantable soft optical waveguides have already achieved significant advancements across biomedical domains, encompassing in situ analytical detection (comprising the monitoring of parameters such as glucose, [34,61,122] blood oxygen, [50,123] etc.), therapy, [124][125][126][127] and optogenetics. [49,128,129]lucose sensors are a crucial component of blood glucose monitoring for diabetic patients.A glucose-sensitive hydrogel fiber has been developed that is capable of quantifying glucose in vivo. [61]Glucose molecules combine with the benzeneboronic acid groups in the hydrogel fiber, thereby altering the diameter of the hydrogel fiber and consequently influencing optical transmittance loss (Figure 7a).The glucose concentration can be quantified by detecting light fluctuation.Utilizing a glucosesensitive sensor, insulin release can be quantitatively triggered when a patient's glucose levels reach unhealthy levels.In addition to glucose detection, Choi et al. leveraged the spectral absorption properties of hemoglobin to implant hydrogel optical fibers into live mice. [50]Two core-clad hydrogel optical fibers were inserted subcutaneously in the mice, with one fiber for light transmission and the other for light collection (Figure 7b).Following the Beer-Lambert law, changes in light intensity were utilized to calculate the relative concentrations of oxygenated and deoxygenated hemoglobin.This design offers potential applications in blood oxygen level monitoring and medical imaging.Furthermore, Reproduced with permission. [61]Copyright 2017, Wiley-VCH.b) Implantation of hydrogel fibers in an anesthetized mouse for monitoring blood oxygenation levels.Reproduced with permission. [50]Copyright 2015, Wiley-VCH.c) Cumulative release of naproxen from the azobenzene grafted carboxymethyl cellulose hydrogels at 37 °C.Reproduced with permission. [137]Copyright 2020, Elsevier.d) Orbital-waveguide insertion for equatorial scleral cross-linking.Reproduced with permission. [139]Copyright 2019, Association for Research in Vision and Ophthalmology.e) Schematic diagram of an electrode array coupled with hydrogel optical fiber.Reproduced with permission. [140]Copyright 2018, Wiley-VCH.
An increasing number of implantable soft optical waveguides are being developed for biomedical treatments. [38,136]Employing intelligent optical waveguide sensing or actuation contributes to precise drug release, thereby enhancing therapeutic efficacy.Kim et al. designed carboxymethyl cellulose hydrogels for controllable drug delivery. [137]The hydrogels were prepared through host-guest complexation, involving azo-tethered carboxymethyl cellulose and β-CD dimers linked by disulfide bonds, with agarose serving as the structural support.The hydrogels have the capability to release up to 80% of the drug within 3 h using either UV light or a reducing agent (Figure 7c).Furthermore, softwaveguide actuators can be engineered as stimuli-responsive gates to allow precise control over drug release.This is achieved through the utilization of photothermally responsive polymer hydrogel films. [138]The polymer undergoes reversible conformational changes when exposed to near-infrared (NIR) light, acting as an on-off gate to control drug release.This controlled release mechanism was demonstrated for drugs like doxorubicin and lysozyme, achieving high bioavailability and significant cell viability reduction in vitro after 5 min of IR light exposure.In addition to drug delivery, flexible soft optical waveguides have also been investigated for myopia treatment.Kwok et al. designed a core-cladded structure waveguide to selectively deliver light and harden sclera without damaging the surrounding tissue. [139]he proposed optical waveguide, combined with scleral collagen cross-linking techniques, mechanically reinforces scleral tissue and prevents subsequent axial length changes (Figure 7d).This offers a solution for controlling its progression in cases of high myopia, thereby reducing the risks associated with sight-threatening conditions such as glaucoma and retinal detachment.
Another promising application for soft optical waveguides is optogenetics, aiming to overcome the limitations of light penetration into deep tissues within biological organisms as well as to avoid tissue damage.Wang et al. fabricated an alginate-polyacrylamide hydrogel optical fiber with a low Young's modulus, high elasticity, and low light propagation loss (Figure 7e). [140]he manufactured fibers were implanted into the brains of mice, and at the end of a 4 week period post-implantation, there were no notable alterations in the fibers' extensibility and optical conductivity.Additionally, there was no substantial neuronal loss observed in the vicinity of the hydrogel fibers.These studies will contribute to the exploration of novel soft optical waveguide materials for optogenetics.Soft actuators have the advantage of being minimally invasive while being able to make contact and movement with delicate tissues.A two-axis nanopositioner has been demonstrated based on a simple graphene-elastomer model system. [141]Through the entropy elasticity of pre-strained graphene/elastomer composite materials, precise and controllable mechanical motion was generated.The embedded graphene nanoplatelets induced efficient light absorption, followed by energy propagation to adjacent polymer chains.The nanopositioner resolution is %120 nm.The purposed nanopositioner is well suited for use in systems requiring precise positioning control/operation, such as probe stations and optical microscopes.Current research is mostly at the technology validation stage, and the functionality is not diverse enough compared to rigid probes.Therefore, new design, processing, and application solutions for soft waveguide are needed.

Wearable Physiological Devices
Soft optical waveguides have the capacity to transduce mechanical stimuli into detectable variations in optical signals.Building upon this capability, relevant wearable devices can be designed for the analysis and monitoring of human body movements.Soft-optical-waveguide sensors and actuators offer a multitude of advantages in physiological activities monitoring. [81,83,142]irst, their low operational noise ensures a peaceful monitoring experience without causing disturbance to the surrounding environment.Second, soft optical waveguides demonstrate high energy efficiency and precise motion control, facilitating rapid responsiveness to user commands in wearable devices without interference with other electronic devices.
Breathing rate (BR) and heart rate (HR) stand as important physiological parameters within wearable devices, offering not only insights into users' health and physiological states but also enabling tailored health management and precise exercise monitoring.An intensity-based polymer optical fiber sensor has been fabricated for monitoring both BR and HR during dynamic movements. [143]The analysis of sensor response is conducted in the frequency domain.Through the integration of a Butterworth filter, the sensors exhibited errors of less than 2 breaths per minute for BR and 4 beats per minute for HR, even when users engaged in periodic movements.Guo et al. fabricated highly stretchable optical strain sensors based on dye-doped PDMS fibers. [81]The fundamental principle underlying this strain sensor is founded on the absorption characteristics of dye molecules, in accordance with the Beer-Lambert law.The optical fiber's attenuation increases linearly with its length when subjected to stretching.By attaching the sensor to the volunteers' necks for testing muscle movement monitoring, it demonstrated the ability to perceive sound vibrations generated during speech and pressure related to inhalation and exhalation (Figure 8a).The sensor exhibited strong repeatability and responsiveness.
In addition to these small range-of-motion examples, almost all body movements, such as finger movement and joint movement, can be monitored with high level of comfort by utilizing soft optical waveguides with a suitable operating range. [144,145]eber et al. used a continuous and scalable melt-flow process to coextrude two thermoplastic elastomers, thereby form stretchable step-index optical fibers. [83]This intensity-based sensor is capable of following the continuous motion of five fingers and knee bending angle in real time (Figure 8b).In addition, the sensor can withstand up to 300% reversible strain while directing light.This device could improve the quantitative assessment of human movement in rehabilitation, sports, and anywhere else where large deformations need to be reliably monitored.
Wearable soft-optical-waveguide technology is not only used for tracking human body movements but has also expanded into fields such as temperature monitoring.Since temperature is another dimension of physical information beyond mechanical perception, mimicking temperature receptors on human skin is critical to expanding the functionality of wearable devices.Guo et al. developed a temperature-sensitive, wearable and biocompatible step-index PDMS optical fiber incorporated with upconversion nanoparticles (UCNPs). [64]UCNPs offer dual-wavelength thermosensitive upconversion emission, enabling proportional temperature sensing through NIR excitation.The sensor exhibited rapid response within the temperature range of 25-70 °C, achieving a detection limit of AE0.3 °C.The proposed sensor has a wide range of applications in real-time monitoring of human body temperature and thermal activities, including the mouth, skin, and nasal breathing (Figure 8c).Soft-optical-waveguide-based sensors offer a secure, reliable, and innovative approach for wearable devices.Moreover, numerous applications of immersive artificial intelligence-based technologies are closely associated with human body movements.This synergy holds the potential to drive the advancement of optical waveguides in entertainment systems and immersive technologies.

Soft Robotics
Robotics technology is a critical field in the era of humanmachine interactions.[155][156][157] McCandless et al. demonstrated a soft robot equipped with multimodal sensing capabilities for 3D shape perception and touch recognition through adjustable soft optical sensors. [158]Laser micromachining techniques are employed to modify the sensor's surface roughness, allowing soft optical sensors to generate asymmetric sensor responses when bent in different directions.They demonstrated the system's ability to accurately track its position and shape during tasks.This includes moving to an object, picking it up, and relocating it.These findings highlight the precision and adaptability of the soft robotic system in performing tasks that require complex movements and interactions with its environment.A prosthetic finger has been developed with flexible chromatic optical waveguides implanted at the fingertip. [159]y detecting the change in the optical loss of the multichannel optical waveguide, the developed prosthetic finger successfully recognized various written letters and Braille characters, and accurately determined the instantaneous sliding conditions of manipulated objects (Figure 9a).This design can be applied to smart prosthetics and robots, offering a new avenue for precise fingertip haptic sensing and manipulation.
Soft robots constructed using soft optical waveguides have showcased significant versatility in challenging environments.Mo et al. proposed a new multidirectional external perception soft actuator based on two flexible-optical-waveguide sensors. [160]he soft actuator can recognize 12 contact positions based on the light gradient boosting machine model, achieving a remarkable recognition accuracy of 99.82%.According to the contact location feedback, teleoperation can be more efficiently completed in unknown underwater environments (Figure 9b).It enriches the ability of soft actuators to sense external stimuli and build their perceptual processing system like the human hand.Meantime, characteristics of living organisms such as light repellency, deformability, and agility have inspired researchers to incorporate nature-inspired elements into robotic structures to enable robots to adapt to unpredictable and unknown environments. [161]Xiang et al. fabricated an actuator which is composed of a porous silicone elastomer and graphene oxide. [162]t exhibits a reversible and well-integrated response to NIR light due to the photothermal-induced contractile stress in the actuation film.This promotes the generation of cyclical and rapid  [81] Copyright 2017, The Optical Society.b) Integration of fiber-based sensors into knee brace and glove for activity monitoring.Reproduced with permission. [83]Copyright 2018, Wiley-VCH.c) Sensor response when attached to the skin or embedded in the mouth of a volunteer.Reproduced with permission. [64]Copyright 2019, Wiley-VCH.locomotion upon NIR light being switched on and off, such as bending in air and crawling in liquid.The purposed actuator can jump from liquid medium to air with 400 ms response time, a maximum speed of 2 m s À1 , and a height of 14.3 cm (Figure 9c).These findings hold great potential for the application of bioinspired actuators in various fields, including microrobotics, sensor development, and advanced locomotion systems.

Soft-Waveguide Outlook
This article reviews the recent research progress in different fields based on soft optical waveguides, including the materials, fabrication, characterization, sensing, and actuation strategies of soft optical waveguides.Good sensing and actuation performance, simple system construction, and multifunctional response capabilities herald a bright future for soft waveguide.Although a variety of promising soft-waveguide sensors have been soft waveguides do not yet offer comparable multimodal sensing capabilities compared to natural organisms.For example, human skin has about 10 000 mechano-sensitive neurons per hand, which can provide multiple sensing capabilities for mechanical, temperature, and chemical stimuli.Therefore, to make the soft-waveguide-based sensor closer to the real sensing capability, the biochemical sensing (temperature, glucose, etc.) and physical sensing (stretching, bending, etc.) of the soft waveguide should be combined.This requires the fusion of different materials and the improvement in fabrication.The integration of multimodal sensors is akin to the comprehensive processing of environmental information by the human brain.Considering that multimodal sensing increases the amount of information to be processed, it is crucial to ensure that the system processes data quickly, and filter invalid and incorrect information, thus ensuring the system makes accurate decisions.Therefore, algorithms need further optimization.
Currently, sensing applications based on soft waveguides are primarily accomplished using neural networks or a combination of numerical simulation and neural networks.These models are effective in solving simple multimodal deformation classification, such as bending, stretching, and pressing.However, in fields such as augmented reality, higher demands are placed on sensors, which require higher resolution and faster response speeds to address challenges arising from free deformation, uncertainty, and noise in complex sensing modes.For instance, unexpected deformation of soft waveguides may cause errors in output information.It is necessary to enhance the robustness of the algorithm to accurately identify user behavior.Similarly, realtime analysis algorithms can be developed to enhance the learning ability of soft waveguides.Similar to how humans adapt to changing environments, the utilization of real-time analysis algorithms has the potential to enhance the learning capabilities of soft waveguides.Moreover, soft waveguides can combine with visualization algorithms to yield enhanced sensing and analytical capabilities.Lastly, ML can be applied to the transfer of soft optical waveguides, i.e., applying learned models to new waveguides or environments.This transfer can improve the adaptability and flexibility of optical waveguides and reduce development costs and time.By transferring already trained models to new sensors or environments, the functionality of soft waveguides can be realized more quickly, and the need for on-site experiments and data collection can be reduced.
Despite the impressive advances in various aspects of soft waveguide, many of the achievements reported in the literature are proof of principle or laboratory environment-based studies.Compared to traditional glass fibers, soft optical waveguides still exhibit significant intrinsic inherent optical loss. [163]It is necessary to optimize the soft-waveguide structure and fabricating process to reduce inherent optical transmission losses. [164]he fabrication of specialized soft-waveguide structures, such  [159] Copyright 2023, Royal Society of Chemistry.b) The whole grasping process of the soft hand in an underwater environment.Reproduced with permission. [160]Copyright 2023, Wiley-VCH.c) Jumping of the PDMS/graphene oxide film from ethanol solution to air.Reproduced with permission. [162]Copyright 2022, American Chemical Society.
as gratings, requires high-precision machining.The grating period is usually on the order of micrometers, but for short period gratings, the period is less than 1 μm.The fabrication precision for such fine structures must reach the scale of tens of nanometers or even lower. [165,166]From a fabrication perspective, a challenge in the layered design of soft optical waveguides is the lack of compatibility between manufacturing technology and of different length scales.For example, soft microstructure waveguides may be susceptible to damage during the fabrication process due to other manufacturing techniques.In addition, some previously reported optical waveguides require the combination of rigid materials, which limits the softwaveguide flexibility and biocompatibility.Therefore, the design and manufacturing of soft optical waveguides need to break away from the dependence on rigid materials.Meantime, because soft waveguide needs to continuously undergo deformation or perceive changes in the external environment during use, it is necessary to carry out various tests and optimizations to achieve reliable and long-term stable use.Certain soft optical waveguides are vulnerable to failure, primarily because of the sensitivity of soft materials to environmental fluctuations, mechanical impacts, and punctures.[169] This self-repairing characteristic not only enhances the resilience of soft optical waveguides but also extends their overall service life significantly.The development of self-healing materials has great impact on the practical applications and longevity of soft optical waveguides, ensuring their continued performance and reliability in demanding environments.Biocompatibility, especially in the case of biodegradable implantable optical waveguides, holds exciting potential for clinical applications, particularly in the context of avoiding retrieval operations and recovery procedures. [41,170]Melt drawing techniques have facilitated the widespread utilization of glass fibers.It is imperative to investigate approaches for the large-scale production of soft optical waveguides.The exploration of soft-optical-waveguide materials with exceptional optical performance, amenable to large-scale manufacturing, or the development of efficient short-term curing methods is essential to enable the extensive application of soft optical waveguides.Intelligent functions such as autonomous motion, self-sensing, self-regulation, and self-learning capabilities should be further promoted.The advancement of these functions can greatly facilitate their application as next-generation intelligent soft robots and wearable devices.In conclusion, the intelligent development of soft optical waveguide requires comprehensive research and exploration in terms of technology, algorithms, and application scenarios to realize their wide applications.

Figure 2 .
Figure 2. Optical intensity changes based on deformations of soft waveguides.a) Intensity variation of the silicone-based sensor under dynamic stick-slip situation.Reproduced with permission.[77]Copyright 2022, Wiley.b) Design layout of the polydimethylsiloxane (PDMS) sensor and the position of four subsequent contacts.Reproduced with permission.[78]Copyright 2013, MDPI.c) Performance of multidirectional self-sensing actuators.Reproduced with permission.[68]Copyright 2022, American Association for the Advancement of Science (AAAS).d) Two-chamber-waveguide signal strength during single and combined strain modes.Reproduced with permission.[80]Copyright 2022, IEEE.e) Performance of soft-waveguide sensor for autonomously selfhealing elongation and flexion sensing.Reproduced with permission.[86]Copyright 2022, AAAS.

Figure 3 .
Figure 3. Optical intensity changes based on chemical reactions of soft waveguides.a) Light propagation and thermosensitivity of the temperatureadaptive hydrogel fiber-based optical waveguide.Reproduced with permission.[67]Copyright 2022, Springer Nature.b) Performance of glucose-sensitive hydrogel optical fibers under different time and concentrations.Reproduced with permission.[61]Copyright 2017, Wiley-VCH.

Figure 4 .
Figure 4. Other sensing strategies of soft waveguides.a) Specific wavelength sensing of Hg2 ions based on fluorescent hydrogel waveguide.Reproduced with permission.[90]Copyright 2017, Springer Nature.b) Specific wavelength sensing of Pb 2þ ions based on quantum-dots-doped tapered hydrogel waveguide.Reproduced with permission.[91]Copyright 2018, Springer Nature.c) Contact location and force sensing based on four fiber Bragg gratings embedded in soft material.Reproduced with permission.[95]Copyright 2020, Mary Ann Liebert.d) Wavelength and color sensing of compression and deformation based on PDMS waveguide with grating structure using e-beam lithography.Reproduced with permission.[60]Copyright 2022, MDPI.e) Phase shift sensing of temperature based on gel polymer waveguide.Reproduced with permission.[99]Copyright 2021, American Chemical Society.

Figure 5 .
Figure 5. Algorithms used in soft waveguides.a) Machine-learning-based construction of the sensor inverse function.Reproduced with permission.[95]Copyright 2020, Mary Ann Liebert.b) Design and sensing result of the proposed multifunctional soft sensor.Reproduced with permission.[84]Copyright 2020, American Association for the Advancement of Science.c) Conceptual schematic and simultaneous realization of sensing and perception.Reproduced with permission.[103]Copyright 2022, Springer Nature.

Figure 7 .
Figure 7. Biomedical detection and therapeutics based on soft optical waveguides.a) Implantation of hydrogel optical fibers in porcine tissue.Reproduced with permission.[61]Copyright 2017, Wiley-VCH.b) Implantation of hydrogel fibers in an anesthetized mouse for monitoring blood oxygenation levels.Reproduced with permission.[50]Copyright 2015, Wiley-VCH.c) Cumulative release of naproxen from the azobenzene grafted carboxymethyl cellulose hydrogels at 37 °C.Reproduced with permission.[137]Copyright 2020, Elsevier.d) Orbital-waveguide insertion for equatorial scleral cross-linking.Reproduced with permission.[139]Copyright 2019, Association for Research in Vision and Ophthalmology.e) Schematic diagram of an electrode array coupled with hydrogel optical fiber.Reproduced with permission.[140]Copyright 2018, Wiley-VCH.

Figure 8 .
Figure 8. Wearable physiological monitoring of the human body.a) Sensor response to a volunteer speaking and deep breathing.Reproduced with permission.[81]Copyright 2017, The Optical Society.b) Integration of fiber-based sensors into knee brace and glove for activity monitoring.Reproduced with permission.[83]Copyright 2018, Wiley-VCH.c) Sensor response when attached to the skin or embedded in the mouth of a volunteer.Reproduced with permission.[64]Copyright 2019, Wiley-VCH.

Figure 9 .
Figure 9. Soft optical waveguides in the manufacture of robots.a) Demonstration of prosthetic-finger-based braille recognition.Reproduced with permission.[159]Copyright 2023, Royal Society of Chemistry.b) The whole grasping process of the soft hand in an underwater environment.Reproduced with permission.[160]Copyright 2023, Wiley-VCH.c) Jumping of the PDMS/graphene oxide film from ethanol solution to air.Reproduced with permission.[162]Copyright 2022, American Chemical Society.

Table 1 .
Representative soft optical materials and strategies for waveguide fabrication.