[ ARTICLE ]
The Journal of Korea Robotics Society - Vol. 16, No. 1, pp.1-11
ISSN: 1975-6291 (Print) 2287-3961 (Online)
Print publication date 01 Mar 2021
Received 07 Oct 2020 Revised 29 Oct 2020 Accepted 30 Oct 2020

# Soft Optical Waveguide Sensors Tuned by Reflective Pigmentation for Robotic Applications

Babar Jamil1 ; Youngjin Choi
1PhD Student, Department of Electrical and Electronic Engineering, Hanyang University, ERICA, 15588, Ansan, Korea babar2015@hanyang.ac.kr
로봇 어플리케이션을 위해 반사 색소로 조정된 소프트 광도파로 센서
바바르 자밀1 ; 최영진

Correspondence to: Professor, Corresponding author: Department of Electrical and Electronic Engineering, Hanyang University, ERICA, 15588, Ansan, Korea ( cyj@hanyang.ac.kr)

## Abstract

Soft robotics has attracted a huge amount of interest in the recent decade or so, be it either actuators or sensors. Recently, a soft optical waveguide sensor has proven its effectiveness for various sensing applications such as strain, force, and bending measurements. The operation principle of the waveguide is simple, but the present technology is far too much complex to manufacture the waveguide. The waveguide fails to attract various practical applications in comparison to other types of sensors despite its superior safety and ease working principle. This study pursues to develop the soft sensors based on the optical phenomena so that the waveguide can be easily manufactured and its design can be conducted. Several physical properties of the waveguide are confirmed through the repetitive experiments in the aspects of strain, force, and bending of the waveguide. Finally, the waveguide sensor is embedded inside the actuator to verify the effectiveness of the proposed waveguide as well as to extend the application fields of the waveguide sensor.

## Keywords:

Soft Optical Waveguide, Soft Sensor, Force, Bending Sensor, Lithography

## 1. Introduction

Optics since its discovery has attracted much interest in sensing applications from astrophysics to optical fibers. Narrowing it down to a very specific level, soft optical waveguides are now well-established in recent years, for their use as sensors, and their applications as sensors in soft actuators. The waveguide has a core in a cladding made of fiber in relation to fiber optics [1-4], making use of light propagation through a combination of elastomers having several properties. Likewise, the operation principle of fiber optics, the waveguide can respond to physical deformation due to the intensity variation of photons arriving at the receiver end, and further, it is very sensitive to be used as sensor. The core and its cladding are fabricated using totally different materials with their own respective refractive indices. The waveguides are supposed to be highly responsive towards physical changes applied to themselves [5-9]. These physical deformations can be in the form of external pressures due to forces on an area, bending, or strains.

A soft optical waveguide in the normal physical form is shown in [Fig. 1] and it consists of a core, cladding and a soft body compounded in a soft actuator or surface. The fabrication of waveguides has been carried out as two different methods [5-8], and how to fabricate them easier would be a big challenge. Elastomers have their own different properties as suggested in [5], the refractive index and absorbance characteristics play a pivotal role in making these waveguides work, however, in general, not every material could match these criteria, and thus, the fabrication of the waveguide to be used in actual applications is still a challenge. A different approach to the fabrication of cladding using gold foil for soft optical sensor has been proposed in [6,7]. This approach resulted in achieving the soft optical waveguide with high elasticity thanks to the elastomers with very low stiffness such as Eco-Flex 30, PDMS silicone, etc. Although the waveguide has shown good response properties in terms of pressure and strain, however, the results do not seem to be promising due to the fact that the strain performance is highly nonlinear because of the deterioration of the measured signal, and also it cannot measure the bending.

Basic representation of a soft optical waveguide sensor in a simple form, where LED and PDE are light-emitting diode and photodiode, respectively. LED works as source and PDE as detector

Using the elasticity and flexibility of soft sensor, Eutectic Gallium-Indium (EGaIn) based soft sensor has been suggested in [10,11]. This sensor is usually fabricated using the conductive liquid filled in the micro-channels of the soft sensor. This approach makes it possible to integrate these liquid conductors to be fabricated/injected within the actuator body resulting in an in-body integration. Despite the possibility to be integrated into the soft actuator, however, there are several challenges associated with these types of sensors. First, the electrical resistance of the sensor is very low, e.g. a few Ohms, and it makes the sensor sensitive to small change and moreover it makes the resolution be very low. Second, the physical stress due to the strain makes micro-channels of the sensor be deteriorated rapidly and further it results in liquid leakage between micro-channels of the sensor. Third, the liquid leakage issue makes the sensor system be dangerous when the soft sensor is working together with other sensitive electronics or electrical equipment.

The potential for optical waveguides sensors is immense in that the LED with driving voltage 5[V] and current 10[mA] would generate a few billions of photons. It is able not only to transfer a huge amount of information but also to bring a high resolution that cannot be matched with any other type of sensors such as an EGaIn-based sensor. The soft optical waveguide can be used not only for one-dimensional deformation but also for three-dimensional changes. It might be able to make the waveguide be integrated into many applications. If the soft actuator including the self-sensing capacity of the optical waveguide might result in low-cost robots. However, currently available fabrication methods prevent from various application extensions of these sensors. Also, in the case of soft actuators or robotics, there can be specific physical constraints imposed by the applications such as the requirement for either high elasticity and low stiffness or high flexibility and low elasticity, and so on [12-16]. These could be challenging works and important issues for almost every self-sensing actuators in soft robotics.

The usefulness of any sensor can be characterized by several properties such as its sensitivity, resolution, physical deformation and response speed. The aim of the paper is to find a new method in the fabrication of soft optical waveguides that can work as soft sensors, so as to improve the performance of soft optics-based sensors and make it possible to fabricate these sensors with application-specific properties. Another aim of this study is to increase the possibility of application for soft optical waveguide sensors and to resolve the shortcomings during fabrication. We use an infrared (IR) spectrum because the frequency ranges of IR signal are independent of ambient light sources, it helps keep the integrity of the spectrum in the waveguide providing accurate and uninterrupted measurements. In the subsequent sections, the calibration as well as experimental setups for characterization are presented to measure the characteristics of the proposed sensor. This paper is organized as follows. The design, fabrication, and materials are presented in section II, and section III suggests several characteristics of the proposed sensor with calibration setup being used. An application of the proposed sensor is presented with the experimental results and the importance of the proposed sensor in section IV, and finally, section V concludes the paper and discusses possible future applications.

## 2. Materials and Methods

Soft actuators and sensors are most commonly manufactured using a combination of elastomers and flexible materials. When being manufactured using elastomers, soft lithography is mainly used [17,18], but it is noted that the optical properties (e.g., absorbance, reflection, and refraction) of the elastomers may be varied. It is known that many elastomers with high elasticity are found to have low absorbance and high penetration for optical radiation interacting to its surface. Soft optical waveguides are manufactured by several methods such as FBG (Fiber Bragg Grating)-based, pure elastomer-based or other hybrid techniques including gold foil-covered cladding for soft optical waveguide core. These waveguides then result in signal power-loss according to deformations applied to themselves such as bending, pressing, and strain. Thanks to these deformations, the soft optical waveguides can potentially be used as soft sensors for various applications, although their current manufacturing schemes bring some limitations.

Let us consider an overall physical phenomenon caused inside an optical sensor as shown in [Fig. 2], in which a photon radiation emitting source will spread optical radiations in the material covering it. Depending on the propagation properties such as refractive index and absorbance, photons will be spread in every direction. It means that some photons would travel straight while some spread all around due to the difference in material properties. However, the refractive photons will not enter back into the waveguide. If there are reflective crystals present in the waveguide, these would reflect the photons coming in the cladding towards waveguide core and in the direction of the detector, depending on the combination of the direction of radiation flow, reflection, and refraction. For this case, if the reflection property of an elastomer can be improved, then the overall high optical flow can be achieved from the source to the detector end. The optical radiation reflected at the detector end will highly be dependent on the relation defined in the following equation:

Illustration of the physical phenomenon caused inside the pigmented soft optical waveguide sensor

 ${P}_{de tector } \alpha \left(\frac{{P}_{g}}{{P}_{e}A}\right)$ (1)

where $\frac{{P}_{g}}{{P}_{e}}$ represents a ratio of the particle density of pigments to elastomer particles in the area A to which photon radiation collides. The relation of Eq. (1) expects that the quantity ratio $\frac{{P}_{g}}{{P}_{e}A}$ dealing with the reflective property within the cladding over area A can effectively increase the amount of optical radiation detected at the receiver end. In the case of reflective crystals, pigments have the property to reflect light when light particles contact themselves. Silicone pigments are one of the choices of reflective crystals to increase reflective property on a specific wavelength for soft elastomers. This means a white light would result in a specific wavelength reflection in a material. When soft optical waveguides are considered, optical radiation or light is propagated from the source towards detector end, and also, when the light is propagated from the source material into the cladding, the radiation would spread around and may lead to emission and absorbance within cladding area. In our experiments, we try to make optical radiation be reflected towards the core area so that the propagation is directed towards the detector end and thus the overall radiation transmission happens towards detector end.

To make these soft optical waveguides work, in general, the elastomer of a specific mechanical property is chosen, and the specific density of pigment crystals are added into the solution as shown in [Fig. 3]. The fabrication method suggested in the figure is in general different from the complex methods proposed in other researches. Also, the soft optical waveguide can be tuned for mechanical properties as well as propagation properties. Mechanical properties able to be tuned are stiffness and flexibility, the stiffness property directly influences the strain capability of these sensors, meaning either these are stretchable and bendable or only capable of one deformation and so forth. The propagation properties can be tuned by the amount and type of reflective crystals added into the elastomer for the fabrication of cladding or waveguide, the crystals can be of any type if the physical properties after waveguide fabrication are intact.

Making process for solution mixed pigment crystals and elastomers

The proposed new soft waveguide is fabricated using elastomers of desired physical properties (stiffness/flexibility), in addition, there are no restrictions to have the specific refractive or reflective properties for materials in use [19], except that the core part and the cladding should have low absorbance and loss, respectively. In the paper, we fabricate a new soft optical sensor with almost similar refractive indices, by using Eco-Flex 30 for waveguide and Vyta Flex 20 (1.43) for the core. The reason why we choose Eco-Flex 30 is because of its high elasticity having the ability to endure the strain of about ten times its original size. And thus, it can be a prime candidate for higher strain and lower stiffness applications. The optical reflective silicone pigments are added with the ratio 1/65. The entire procedures for the fabrication of these silicone waveguides are illustrated in [Fig. 4]. Eco-Flex 30 is prepared and the silicone pigments are added to the ready solution, after adding pigments (the ratio of pigment to elastomer can be varied), the solution is mixed together at 2000 RPM (revolutions per minute) and, after the mixing process is done, the mixture is degassed using a vacuum chamber, and the degassed mixture is poured on the mold for the waveguide. This mold is then put into the oven for 45 minutes, and then LED and PDE side molds are attached to both ends of the fabricated waveguide.

The entire fabrication process for the proposed new type of soft optical sensors

Although the core of the waveguide can be fabricated with any material without reflection property, for our study, Eco-Flex 20 is used to fabricate the core part for the soft optical waveguide, last but not least, careful attention is required to guaranteed no air bubbles left in the core part. This is done by the proper vacuum of the mixed Vyta Flex 20 solution. The precise bubble-free needle to lay down the core part of the waveguide is used, and the half-prepared waveguide is put in the oven again for an hour. After this process, the remaining step involves fabricating the uncovered part of the waveguide and completing the procedure.

The physical design of these soft waveguides is changed by adding an inex tensible layer used to connect them to other instruments or parts inside or outside actuators. A textile layer with Eco-Flex 30 is used to make inextensible connectors for these sensors because the fabric or any other inextensible layer is flexible but not elastic. Prototype sensors made using the proposed method are compared to the existing method [19], however, in our case using Mold Max 30 in terms of the physical characteristics (elasticity) as well as to prove the effectiveness of the proposed method. Three prototypes are shown in [Fig. 5], with two different pigments of blue and red colors.

Three different soft optical sensors for the property comparison, where Eco-Flex 30 plus blue and red pigmentations are fabricated by the proposed method, and Mold Max 30 plus red pigmentation is made by the method suggested in [19]

## 3. Characterization of Optical Sensors

Soft actuators and sensors involve hyperelastic materials and nonlinear deformations, this makes it very hard for their mathematical analysis. As an alternative, experiment-based models play an important role in characterizing soft actuators and sensors. This is usually done by plotting input-output curves, strain-output curves and so on. There are three properties associated with these soft optical waveguides and, in our study, two types of experimental setups are used to study the characteristics of these waveguides. Strain, bending and pressing characteristics are presented here for these optical waveguides. The signal is measured and plotted as a relative power loss from an initial value of the signal received at the detector end. In general, the signals are acquired using a data acquisition board at 5[kHz] sampling frequency. For clarity purposes, the voltage is measured across the resistor attached to the PDE end and it is used to calculate the power loss as follow:

 ${P}_{dB} \alpha 20 l {og}_{10} \left|\frac{{V}_{{R}_{C}}}{{V}_{{R}_{i}}}\right|$ (2)

where VRi and VRc represent the initial voltage-drop without any deformation and the voltage-drop when being deformed, respectively.

### 3.1 Elongation Characteristics:

The relationship between strain to power loss is measured as elongation characteristics and, to characterize this property of these soft sensors, a linear actuator that can keep well track of the desired displacement is used with high accuracy. For this experiment, strain is applied horizontally along the body of soft optical waveguides. This process is carried out by applying strains from a linear system that is controlled using a driving mechanism. The signal are collected using data acquisition device (DAQ NI-6216) at a sampling rate of 5KHz. When the linear strain is considered in the soft optical waveguides, the Beer-Lambert equation is used to relate the power loss to the strain as follow:

 ${Ab}_{m}=ecL=ec\epsilon {L}_{0}+B$ (3)

where Abm is a total optical absorbance (optical density) in the core material, and the strain is defined as $\epsilon =\frac{L-{L}_{0}}{{L}_{0}}$. This property of the core material of soft optical waveguide depends on e the absorbtivity of dye, c the concentration of dye in solution, and L the length of light path [cm]. Also, B=ecL0 denotes a constant baseline absorbance for the core material and L0 the initial length of light path. Since zero absorbance at some wavelength implies that none of the light of the particular wavelength has been absorbed, the absorbance can be defined via the incident intensity I0 and the transmitted intensity I as follow:

 ${{Ab}_{m}=lo g}_{10}\left|\frac{{I}_{0}}{I}\right|$ (4)

Since the optical absorbance in a medium can be calculated using Eq. (3), which results in a linear relationship when only strain is applied to the waveguide, however, the same phenomena can be equally modeled by a simple logarithmic equation described in Eq. (4).

The mathematical relation of Eq. (3) can be used to characterize the curve between strain and power loss. For the information about optical signal propagated to the detector end through the core part of these optical waveguides, the optical power loss is calculated according to the changing length of the core area. Since both (red and blue) pigmented waveguides were manufactured with their own physical characteristics, the result might differ according to the pigmentation level and uniformity of pigments in the cladding and other factors. Since these must be very important cases to study, we present the important characteristics of these waveguides. The strain is applied to the waveguide using the calibration setup shown in [Fig. 6], in which the moving frame shown in the figure applies the strain to the waveguide by about 75 micrometers. During the elongation characterization, a strain of over 75% was applied to the optical waveguides, and further, a repeated strain was applied to check the repeatability of these waveguides as well. Over 100 times repeated strain was applied with different magnitude and the data were collected using the acquisition device. The measured data points for strain and resulting power loss are shown in [Fig. 7]. This power loss is calculated using the relative power loss formula, and the relative power loss uses the ratio between the initial baseline signal without strain and the signal when strain is applied. The blue dots show data points collected during strain applied, these data points show some spread for the same value due to the noises and variation in the signal. This variation remains 5% of the mean value. A regression model is calculated using a first-order polynomial model, and the regression fit is shown in [Fig. 7] with a red line. The graph indicates a linear signal variation against strain applied, this linear power loss is a very important property of these waveguides, and this apparent linearity suggests that these waveguides can be used for strain applications with relative ease. The overall loss rate can be adjusted by improving the fabrication processes such as removing impurities.

Experimental setup for elongation characterization of soft optical waveguide

Characteristic curve of average signal power loss in [dB] unit according to the strain variation ϵ, where the experiments are repeated several times

### 3.2 Force / Pressure Characteristics:

One of the advantages of these soft optical waveguides is that they can show the characteristics of power loss according to the increase of the applied pressure. The pressure implies the contact force per unit area when pneumatic/hydraulic pressures are exerted on the surface. For this study, the force was applied upto several newtons on the surface of the sensor and the sensor signal variations were measured using data acquisition device (NI-6216), to have a reference measurement signal, a contact tip was designed and used during contact force application, the contact tip in this case was equipped with aloadcell (Bongshin Co. with a rated capacity of 50N, the entire contact tip was printed using 3-D printer (Stratasys). The square-shaped contact is designed and manufactured using the 3-D printer (Stratasys) with a contact area of 16[mm2]. Analog amplifier (Honeywell UV-10) is used to amplify the signal from the waveguide as well as to reduce the noise effect. Data acquisition device (NI-6216) makes several signals of the loadcell and waveguide be sampled and digitized with 5[kHz] frequency.

An electrical current control scheme is implemented for the linear motor for monitoring the force signal from a reference loadcell. PID controller is implemented to adjust the force of the linear motor, ultimately, in order for applying accurate force to the surface of the waveguide. During the experiments, different waveforms of forces are applied (at lower frequencies) perpendicular to the surface of the waveguide. The power loss of the waveguide according to the applied force variation is shown in [Fig. 8], including the original data and its regression fit. The resulting graph shows an important feature of the waveguide in that the initial power loss is slightly nonlinear until it arrives at 1[N], it becomes linear in the range of 1 ~ 3[N], and it becomes nonlinear again until it arrives at the complete loss of signal at the detector end. Since the contact force was repeatedly applied to the surface of the waveguide over a long time, the waveguide has very good repeatability.

Characteristic curve of average signal power loss according to the contact force variation in [N], where the contact area of the pigmented soft optical waveguide is 16[mm2]

### 3.3 Bending Characteristics:

To characterize the bending property, the waveguide is fixed (attached) to a bending surface which has the ability to provide the bending when the linear actuator moves. Both ends of the bending surface are attached to the moving frame and the fixed frame as shown in [Fig. 9], where the linear actuator moves until the full bending of the surface is completed, and then it returns after the maximum bending towards the maximum flattening of the waveguide. In [Fig. 9], the graph shows the signal change when the red-pigmented waveguide is under the repetitive bending motions. Moreover, the characteristics of power loss to the repetitive bending/releasing is shown in [Fig. 10], where it shows very smooth but nonlinear curve under the consecutive phase changes between the bending and releasing of the waveguide. As another experiment, the bending shape is changed in such a way to add one more bending, ultimately, to observe the variations of the characteristics. The experimental results are shown in [Fig. 11]. The responses in [Fig. 10] and [Fig. 12] have two different points in terms of the measured data characteristics, in the first, the single point bending has resulted in a response curve shown in [Fig. 10], where the characteristic curve shows the power loss of about 22[dB], and it means a rate of 0.12[dB] loss per m-1 curvature. In [Fig. 11], when two-point bending was applied, the characteristic curve shows a higher loss rate, which is institutively related to the physical change. The loss rate, in this case, was 0.153[dB] per m-1 curvature, which is higher than a single point bending with the maximum power loss of 30[dB]. The other property is the difference in the shape of the loss curve, which was different for both single-point bending and two-point bending.

Experimental setup for bending characterization of soft optical waveguides, where time response of the waveguide bending is shown in the figure

Characteristic curve between the power loss measured at the detector and the bending of waveguide

Characteristic curve between the power loss and two-point bending of waveguide

Real time responses of the soft optical waveguide, (a) when arbitrary strain is applied, (b) when arbitrary contact force is applied

Till now, we have shown several properties for practical sensor use in terms of the power loss of the signal. Furthermore, its verification is extended to the real-time responses of the optical waveguides. For this purpose, two different experimental configurations are chosen, such as when the external strains are applied and when the external contact forces are applied. For arbitrarily given strain, the data are measured from the waveguide in response to the strain externally applied and the regression model is used to estimate the strain. Also, the true strain applied by the linear actuator is measured by the Arduino controller. Both signals are compared together to show the performance of the waveguide, as shown in [Fig. 12(a)]. In a similar way, the data are measured from the waveguide in response to the force externally applied and the regression model is used to estimate the force. Also, the true force is measured using the loadcell installed inside the experimental setup. The real-time force data measured from the waveguide and the data measured from loadcell are together plotted in [Fig. 12(b)]. Now, we could know that the strain and force measured from the waveguide were well matched with the actual values. Especially, in the case of force sensing application, the measurement might have a little bit variance due to high nonlinearity, but it is enough to warrant a good performance thanks to the superior sensitivity and resolution.

### 3.4 Physical Characteristics:

The main advantage of the proposed waveguide is not limited to the prototypes proposed, but it could be applied to the different types and designs of waveguides that have various physical characteristics. To explain the advantages of the proposed method, two types of optical waveguides were fabricated, one of the both uses Mold Max 30 and another uses Eco-Flex 30, the fabrication of this sensor waveguide is followed in the same method as explained in the fabrication section. The physical dimensions of the optical waveguides are the same each other such as the sensor length of 100[mm], the core length 55[mm] and its width 3.5[mm].

The strain (or elongation) tests for the waveguides are conducted to show the strain capabilities of these waveguides against external loads attached to the free end of the waveguides as shown in [Fig. 13]. [Fig. 13(a)] shows both the waveguides have the same lengths each other, although the subtle difference could be occurred due to the weights from the connectors. In [Fig. 13(c)], however, we can confirm the big difference of physical strains against two different loads, approximately 0.5[kg] is required to make 40% strain in the case of Mold Max 30 based waveguide, while just 0.2[kg] resulted in over 100% strain in Eco-Flex 30 based waveguide. This difference of characteristics shows how waveguides can be fabricated with different elastic modulus, this elastic modulus will depend on the properties of the core and cladding elastomers. Such as in the case of higher stiffness elastomers will result in an overall highly stiff waveguide and vice versa. It is also mentioned here that the pigmentation itself does not affect the stiffness of the elastic modulus of these waveguides. Therefore, the color of pigmentation added during fabrication will not affect the physical characteristics of these waveguides.

Comparisons of strain characteristics for Mold-Max based waveguide and Eco-Flex based waveguide, in which (a) no load condition, (b) 0.5[kg] is attached to Mold-Max based waveguide, (c) 0.2[kg] is attached to the Eco-Flex based waveguide

## 4. Application to Actuators

There are a number of application areas for the soft optical waveguides as sensors, especially, for the measurements of elongation, force, and bending. The pneumatic artificial muscle actuator is one of the famous types of soft actuators thanks to its high force to weight ratio for some industrial purpose as well as human assistive applications. It is actuated using the pneumatic pressure [16], [20-24] which makes the actuator contract. The contraction of the pneumatic actuator is usually ranged to 25% (commercially available) or up to 40% (research purpose). Since the pneumatic actuator is mostly used to reduce space and weight of the structure being used, it is usually desirable to integrate the sensors inside the actuator rather than to use mechanical sensors outside the actuator due to the size and weight of mechanical sensors. In this study, we use the soft optical waveguide embedded inside the actuator body. Since the displacement or the strain will appear as negative numbers, the soft strain sensor should have a pre-strained physical form to allow the release of the pre-strain. The pre-strain would result in stress, yielding the contraction force in the sensor. To keep the contraction force at minimum, a high elastic sensor body is desired. For this purpose, the Eco-Flex 30 based pigmented soft optical sensor will be integrated inside the pneumatic actuator.

To integrate this soft optical sensor (waveguide), we have used the multi-step processes that are involved with several different steps including the design of the rigid-end of the pneumatic actuator to make it possible for connection and pre-strain stability of the sensor. The design is shown in [Fig. 14] and the fabrication steps of the pneumatic actuator are composed of first fabricating a hallow bladder using hyperelastic elastomer (Dragon Skin 30), second printing the rigid-ends for pneumatic actuator using a 3D printing device (Stratasys), third attaching the soft optical waveguide (sensor) to the rigid- ends, fourth straining soft optical waveguides, fifth attaching these to the ends of the soft bladder, sixth covering the structure in the braided mesh, and finally completing the actuator.

The design concept of a pneumatic actuator having high-elastic soft optical waveguide, where it is fabricated using the proposed method, and the strain arrives at about 70% of the sensor's original length without significant stress

The soft optical waveguide is pre-stretched about 70% of its original length 50[mm] and then fabricated into the body of PAM actuator. This approach is necessary for measuring the contraction of actuator which has negative direction to optical waveguide strain. The amount of pre-stretching may vary depending on the maximum contraction in the actuator, therefore, when actuators are fabricated with the optical waveguide integrated sensors, the dimensions of both actuator and sensor must be considered accordingly. The pneumatic actuator used in our studies has a longitudinal length of 100[mm] with a contraction ratio of about 25% and the inner diameter of the actuator is about 16[mm]. The actuator body with its rigid connections has a weight of 0.08[Kg]. The actuator has an inward stress of 1 [N/m]. The PAM actuator with the sensor integrated inside was attached vertically with weights hanging at the moving end of the actuator, A sampling device (NI-6216) was used to capture the signal from the sensor as well as the pressure input from the actuator.

To check the ability of the high strain sensor to measure contraction of actuator, a load of 0.45[kg] was also attached to the actuator and the pressure was applied in the range of 0 ~ 150[kPa]. The data during these experiments were measured using the data acquisition device as discussed in the section before. During this process, the NI-6216 DAQ device samples the sensor signal at 5[kHz] rate, the contraction of PAM actuators in terms of [mm] displacement of loads was also measured. The results data points are plotted for [dB] power loss against the different displacement of the load. The graph of [Fig. 15] shows the characteristic curve for sensorized PAM actuator, the measured data points are represented using dot in the graph while the characteristic curve fit line is also plotted in [Fig. 15]. This characteristic curve shows in [dB] gain rather than [dB] loss, it is due to the fact that the sensor is initially strained to the size of the PAM actuator as shown in [Fig. 14]. This strain is released during the contraction of the PAM actuator, which is negative along the direction of PAM actuator motion, therefore, the strain is released resulting in the [dB] gain at the PDE end signal. The characteristic curve shows a nonlinear response, meaning that the gain rate is high in the range of 0-11[mm] and low in the range of 11-17[mm]. This result shows that the entire contraction of the PAM actuator can be predicted using two linear models representing the high and low gain rate of [dB] power during contraction of the PAM actuator.

The characteristic curve of the power loss of signal while the actuator is contracted, where the maximum contraction is about 16.9[mm]

## 5. Conclusions and Future Works

The soft optical waveguide has shown a good promise in that it could measure strain, force and bending applied to the waveguide. The results of this study showed high repeatability as well as about 100% strain with a small force. The bending characteristic is important in that the waveguide shows different responses according to the bending shape. It is expected to play an important role in the shape recognition of the soft actuator or robot itself in the modeling and control issues. The response to small contact force is another important discovery of these waveguides, this result shows the potential use in tactile sensing for robotic applications. Another important fact in conducting the experiment is to use infrared (IR) sources and detectors. Since the waveguide has been easily disturbed by the visible light, we used IR- based source and detector to maintain the constant performance in outputs between day and night. Another important fact about IR-based signaling is that the signals carry very little noise and thus it is easy to process the signal measured by the acquisition device.

We are soon expecting to investigate the waveguide with a variety of aspects because the waveguide has responded to all the deformations. For example, it is careful that the signal from the waveguide can respond to vibrations and other phenomena able to disturb itself. Also, one of the best things about the waveguide observed was its good repeatability, which is remarkably superior to any other soft sensors used in our experience. These properties should be confirmed for practical use.

## Acknowledgments

This work was supported in part by a National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (2019R1A2C1088375) and in part by the Technology Innovation Program funded by the Korean Ministry of Trade, industry and Energy (20008908), Republic of Korea

## References

• J. Maher, “Waveguide optical modulator,” U.S. Patent 4128299, Dec., 05, 1978, [Online], https://patents.google.com/patent/US4128299, .
• O. J. Schueller, X.-M. Zhao, G. M. Whitesides, S. P. Smith, and M. Prentiss, “Fabrication of Liquid Core Waveguides by Soft Lithography,” Advanced Materials, vol. 11, no. 1, pp. 37-41, 1999, [https://doi.org/10.1002/(SICI)1521-4095(199901)11:1<37::AID-ADMA37>3.0.CO;2-0]
• R. A. Lieberman, L. L. Blyler, and L. G. Cohen, “A distributed fiber optic sensor based on cladding fluorescence,” Journal of Lightwave Technology, vol. 8, no. 2, pp. 212-220, Feb., 1990, [https://doi.org/10.1109/50.47873]
• Y. Shibata, A. Nishimura, S.-I. Niwa, Y. Osawa, and T. Uemiya, “Optical sensors,” U.S. Patent 4750796, 1988, [Online], https://www.freepatentsonline.com/4750796.html, .
• H. Zhao, K. O'Brien, S. Li, and R. F. Shepherd, “Optoelectronically innervated soft prosthetic hand via stretchable optical waveguides,” Science Robotics, vol. 1, no. 1, 2016, [https://doi.org/10.1126/scirobotics.aai7529]
• C. To, T. Hellebrekers, J. Jung, S. J. Yoon and Y.-L. Park, “A Soft Optical Waveguide Coupled With Fiber Optics for Dynamic Pressure and Strain Sensing,” IEEE Robotics and Automation Letters, vol. 3, no. 4, pp. 3821-3827, Oct., 2018, [https://doi.org/10.1109/LRA.2018.2856937]
• C. To, T. L. Hellebrekers, and Y.-L. Park, “Highly stretchable optical sensors for pressure, strain, and curvature measurement,” 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Hamburg, Germany, pp. 5898-5903, 2015, [https://doi.org/10.1109/IROS.2015.7354215]
• B. Jamil, J. Kim, and Y. Choi, “Force sensing fingertip with soft optical waveguides for robotic hands and grippers,” 2018 IEEE International Conference on Soft Robotics, Livorno, Livorno, Italy, pp. 146-151, 2018, [https://doi.org/10.1109/ROBOSOFT.2018.8404911]
• C. Larson, B. Peele, S. Li, S. Robinson, M. Totaro, L. Beccai, B. Mazzolai, and R. Shepherd, “Highly stretchable electroluminescent skin for optical signaling and tactile sensing.” Science, vol. 351, no. 6277, pp. 1071-1074, 2016, [https://doi.org/10.1126/science.aac5082]
• J.-B. Chossat, H.-S. Shin, Y.-L. Park, and V. Duchaine, “Soft Tactile Skin Using an Embedded Ionic Liquid and Tomographic Imaging,” J. Mechanisms Robotics, May, 2015, [https://doi.org/10.1115/1.4029474]
• Y.-L. Park and R. J. Wood, “Smart pneumatic artificial muscle actuator with embedded microfluidic sensing,” IEEE Sensors, pp. 1-4, Nov., 2013, [https://doi.org/10.1109/ICSENS.2013.6688298]
• S. Davis, N. Tsagarakis, J. Canderle, and D. G. Caldwell, “Enhanced modeling and performance in braided pneumatic muscle actuators,” The International Journal of Robotics Research, vol. 22, no. 3, pp. 213-227, 2003, [https://doi.org/10.1177/0278364903022003006]
• J. Misumi, S. Wakimoto, and K. Suzumori, “Experimental investigation of conductive fibers for a smart pneumatic artificial muscle,” 2015 IEEE International Conference on Robotics and Biomimetics (ROBIO), pp. 2335-2340, Zhuhai, China, Dec., 2015, [https://doi.org/10.1109/ROBIO.2015.7419123]
• G. J. Kookootsedes, H. H. Reese, B. I. Gutek, and G. H. Pretzer, “Touch position sensitive optical waveguides,” U.S. Patent 4701017A, Oct., 20, 1987, [Online], https://patents.google.com/patent/US4701017A, /
• C. Chorley, C. Melhuish, T. Pipe, and J. Rossiter, “Development of a tactile sensor based on biologically inspired edge encoding,” International Conference on Advanced Robotics, Munich, Germany, pp. 1-6, 2009, [Online], https://ieeexplore.ieee.org/document/5174720, .
• B. Tondu and P. Lopez, “Modeling and control of McKibben artificial muscle robot actuators,” IEEE Control Systems Magazine, vol. 20, no. 2, pp. 15-38, Apr., 2000, [https://doi.org/10.1109/37.833638]
• F. Ilievski, A. D. Mazzeo, R. F. Shepherd, X. Chen, and G. M. Whitesides, “Soft robotics for chemists.” Angew. Chem. Int. Ed. 50, pp. 1890-1895, 2011, [https://doi.org/10.1002/anie.201006464]
• C. Laschi, J. Rossiter, F. lida, M. Cianchetti, and L. Margheri, Soft Robotics: Trends, Applications and Challenges, Springer, 2017, [https://doi.org/10.1007/978-3-319-46460-2]
• H. Zhao, K. O'Brien, S. Li, and R. F. Shepherd, “Optoelectronically innervated soft prosthetic hand via stretchable optical waveguides,” Science Robotics, 2016, [https://doi.org/10.1126/scirobotics.aai7529]
• W. Felt and D. C. Remy, “Smart braid: Air muscles that measure force and displacement,” 2014 IEEE/RSJ International Conference on Intelligent Robots and Systems, Chicago, IL, USA, pp. 2821-2826, 2014, [https://doi.org/10.1109/IROS.2014.6942949]
• W. Felt, K. Y. Chin, and C. D. Remy, “Contraction sensing with smart braid McKibben muscles,” IEEE/ASME Transactions on Mechatronics, vol. 21, no. 3, pp. 1201-1209, 2016, [https://doi.org/10.1109/TMECH.2015.2493782]
• B. Jamil, S. Lee, and Y. Choi, “Conductive knit-covered pneumatic artificial muscle (k-PAM) actuator,” 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Madrid, Spain, pp. 1476-1481, Oct., 2018, [https://doi.org/10.1109/IROS.2018.8594510]
• L. O. Tiziani, T. W. Cahoon, and F. L. Hammond, “Sensorized pneumatic muscle for force and stiffness control,” 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, Singapore, pp. 5545-5552, Jun., 2017, [https://doi.org/10.1109/ICRA.2017.7989654]
• J. P. King, L. E. Valle, N. Pol, and Y.-L. Park, “Design modeling and control of pneumatic artificial muscles with integrated soft sensing,” 2017 IEEE International Conference on Robotics and Automation (ICRA), Singapore, Singapore, pp. 4985-4990, Jun., 2017, [https://doi.org/10.1109/ICRA.2017.7989580]
바바르 자밀

2012 COMSATS Institute of Information and Technology (공학사)

2015~현재 한양대학교 전자공학과 석박 통합과정

관심분야: Soft robotics, sensors, actuators, control, machine learning

최 영 진

2002 POSTECH 기계공학과 (공학박사)

2002~2005 KIST 지능로봇연구센터 선임연구원

2005~현재 한양대학교 ERICA 전자공학부 교수

관심분야: 로봇제어, 생체신호처리

### [Fig. 1]

Basic representation of a soft optical waveguide sensor in a simple form, where LED and PDE are light-emitting diode and photodiode, respectively. LED works as source and PDE as detector

### [Fig. 2]

Illustration of the physical phenomenon caused inside the pigmented soft optical waveguide sensor

### [Fig. 3]

Making process for solution mixed pigment crystals and elastomers

### [Fig. 4]

The entire fabrication process for the proposed new type of soft optical sensors

### [Fig. 5]

Three different soft optical sensors for the property comparison, where Eco-Flex 30 plus blue and red pigmentations are fabricated by the proposed method, and Mold Max 30 plus red pigmentation is made by the method suggested in [19]

### [Fig. 6]

Experimental setup for elongation characterization of soft optical waveguide

### [Fig. 7]

Characteristic curve of average signal power loss in [dB] unit according to the strain variation ϵ, where the experiments are repeated several times

### [Fig. 8]

Characteristic curve of average signal power loss according to the contact force variation in [N], where the contact area of the pigmented soft optical waveguide is 16[mm2]

### [Fig. 9]

Experimental setup for bending characterization of soft optical waveguides, where time response of the waveguide bending is shown in the figure

### [Fig. 10]

Characteristic curve between the power loss measured at the detector and the bending of waveguide

### [Fig. 11]

Characteristic curve between the power loss and two-point bending of waveguide

### [Fig. 12]

Real time responses of the soft optical waveguide, (a) when arbitrary strain is applied, (b) when arbitrary contact force is applied

### [Fig. 13]

Comparisons of strain characteristics for Mold-Max based waveguide and Eco-Flex based waveguide, in which (a) no load condition, (b) 0.5[kg] is attached to Mold-Max based waveguide, (c) 0.2[kg] is attached to the Eco-Flex based waveguide

### [Fig. 14]

The design concept of a pneumatic actuator having high-elastic soft optical waveguide, where it is fabricated using the proposed method, and the strain arrives at about 70% of the sensor's original length without significant stress

### [Fig. 15]

The characteristic curve of the power loss of signal while the actuator is contracted, where the maximum contraction is about 16.9[mm]