A Lightweight, Soft Wearable Sleeve for Rehabilitation of ForearmPronation and Supination

Se-Hun Park1, Jaehyun Yi1, DongWook Kim1, Youngbin Lee2, Helen Sumin Koo2, and Yong-Lae Park1

Abstract— Forearm disabilities can result from a number ofdiseases or injuries, such as stroke, spinal cord injuries (SPI),and acquired/traumatic brain injuries (ABI/TBI). Patients withforearm disabilities suffer from daily activities and follow-upcare and rehabilitation are critical to recovery of the patients inthis case. However, they have to make personal trips to clinicsor hospitals for follow-up treatments since most of existingdevices available are bulky, heavy and even expensive in general,discouraging the patients from having them home. This paperproposes a pneumatic inflatable rehabilitation device designedwith bio-inspiration. The device is lightweight, safe, and afford-able compared to conventional devices. When the actuators ofthe device are inflated, the contractions of the actuators mimicthe motions of the human muscles of the forearm, assistingpronation and supination. The actuators are made of heat-sealable fabrics using a custom-built patterning system. Thissystem facilitates rapid fabrication of repetitive and complexsealing patterns. Also, the proposed rehabilitation device, whichcan be conjointly used with typical clothing materials, providescomfort for the user by easily conforming the three-dimensionalgeometries of the forearm. To characterize the system, theangles and the torques of the wrist generated by the devicewere measured. The result of simple closed-loop control usinga vision feedback was also presented.

I. INTRODUCTION

Forearm disabilities are often caused by stroke, fracturesand incised/punctured injuries. Having a cup of water andwearing clothes may no longer be simple daily tasks forpatients with forearm disabilities. In order to help the patientsreturn to their daily lives, it is well known that follow-up care is critical [1], and we propose a lightweight, softwearable device for forearm rehabilitation (Fig. 1). Amongseveral rehabilitation processes, robot-assisted arm training iswidely used these days and improves patients’ motor func-tion effectively [2], [3], [4]. A general mechanism of existingrobotic rehabilitation is that a robot exerts assistive/resistiveforces to help the patient recover function of the forearmmuscles while the patient is holding the handle of the roboticdevice [5], [6], [7].

Traditional devices for robot-assisted rehabilitation aremostly made of rigid materials, making the devices costly

This work was supported in part by the National Research Foundation(NRF) Grant (2016R1A5A1938472) and in part by the Convergent Tech-nology R&D Program for Human Augmentation (2018M3C1B8017699) ofNRF funded by the Korean Government (MSIT). (S.-H. Park and J. Yicontributed equally to this work. Corresponding author: Y.-L. Park.)

1S.-H. Park, J. Yi, D. Kim and Y.-L. Park are with the Department ofMechanical and Aerospace Engineering and Soft Robotics Research Center(SRRC), Seoul National University, 08826, Republic of Korea. (E-mail:{snuhun; ljhtheman; shigumchis; ylpark}@snu.ac.kr)

2Y. Lee and H. S. Koo is with the Department of Apparel Design, KonkukUniversity, 05029, Republic of Korea (E-mail: yblee822@naver.com;smkoo@konkuk.ac.kr)

Fig. 1. Design and mechanism of the proposed wearable device showing(a) the arm bands for easy anchoring of the actuators, the muscles and theactuators for (b) pronation and (c) supination, and (d) an actual prototypeon human arm.

and heavy [8]. Since these characteristics can hamper sub-sequent treatments of the patient, there is a compellingneed of devices made of soft materials (e.g. polymer orfabric). With the advantages of i) compliance to unstructuredgeometries (e.g. curved surfaces), ii) reduced chances ofinjuries in case of malfunctioning, iii) cost efficiency, andiv) lightweight and compact form factor. There have alreadybeen approaches of building orthotic devices for rehabilita-tion using soft materials: a soft tendon-driven glove, madeof silicone, assist flexion and extension motions of patientswith hand paralysis [9]. A bio-inspired soft wearable roboticdevice that can help patients recover from neuromusculardiseases by assisting dorsiflexion and plantarflexion of theankle joint [10], a wearable assistive robotic device, madeof bellow actuators for elbow rehabilitation [11].

However, a soft robotic device that assists wrist motionsof pronation/supination has not been much investigated yet.

2019 2nd IEEE International Conference on Soft Robotics (RoboSoft)COEX, Seoul, Korea, April 14-18, 2019

978-1-5386-9259-2/19/$31.00 ©2019 IEEE 636

Fig. 2. (a) Flat pneumatic inflatable actuator unit. (b) Front and back viewsof the forearm bands.

In order to assist pronation/supination of the forearm, theactuators should be wound around the arm, and conventionalactuators with rigid structures significantly reduces the wear-abilty and the effectiveness of rehabilitation. In contrast toconventional actuators, flat pneumatic actuators are ideal forrotational motions of the wrist due to their near-zero initialvolume. Therefore, we propose a lightweight and compactwearable device made of flat soft pneumatic actuators toassist pronation and supination motions of the wrist (Fig. 1).

Flat pneumatic actuators can be fabricated using a heatsealer with a motorized stage [12]. In this work, we built acustom system to easily fabricate the actuator with a complexpattern. The system was characterized to find the parametersfor increased robustness of the actuator.

An arm model was used for the characterization tests.After installing the device to the arm model, the torques ofthe wrist were measusred with varied input pressures. Then,the torque data were used to achieve target angles using anoptical motion sensing system.

II. DESIGN

A. Bio-inspired design

The design inspired by biological musculoskeletal systemwas based on two forearm muscles: the supinator attachedto the lateral epicondyle and the radius, and the pronatorteres attached to the medial epicondyle and the radius. Twoactuators in an antagonist pair perform the same functionof the two muscles depending on the attachment position.The forearm bands on the upper and the lower parts of theforearm anchor the both ends of the actuators (Fig. 1). Theactuator is wound around the forearm and the attachmentposition can be adjusted by using Velcro straps. Dependingon the wound direction of the actuator, the forearm supinatesor pronates when the actuator contracts.

B. Flat pneumatic actuator

While the human muscles are directly attached to rigidstructures (i.e. bones) and covered with skin, the actuators

Fig. 3. Prototype of proposed wearable device install on arm modelshowing the directions of (a) the actuator contraction and (b) the armrotation, respectively.

have to be placed on the skin. For this reason, the proposedartificial muscles must be slim unlike human muscles sincethe compression of bulky actuators cause discomfort to theforearm. Therefore, the flat pneumatic actuator was chosenover McKibben muscle [13]. With the choice of actuators, theanchoring method is also very important since the skin wherethe actuators are attached may stretch and rotate around thebone. Fabric material was selected for actuators to utilizevarious clothing techniques for increased wearability andeasy fabrication.

The proposed wearable device is composed of two sym-metric flat pneumatic inflatable actuators made of heat-sealable fabric. Each actuator was fabricated using a custom-built motorized heat sealing system and has multiple airchambers connected in series. The pneumatic connections(i.e. air passages) between chambers enable simultaneousinflation (or axial contraction) of all the chambers (Fig. 2).The zero-volume air chamber becomes completely flat whendeflated, which is highly useful for wearability [13]. Theinjection of compressed air into the air chambers makes theactuator axially contract with radial expansion like biologicalmuscles (Fig. 3).

C. Device configuration

The range of pronation/supination of the forearm with anextended elbow position is about 160◦ [14]. Therefore, eachactuator should have a similar range of motion. Althoughthe contraction ratio of general pneumatic artificial musclesranges from 10 only up to 30% [15], a relatively largerotation angle can be achieved if the length of the arm(Lf ) is fixed between the two ends of the actuator. Fig. 4shows the drawing of the flat pneumatic actuator assuming

637

Fig. 4. Actuation mechanism of rotation before (light grey area) and after(dark grey area) contraction.

the forearm is a truncated cone. The blue area with a smallradius indicates the wrist and the yellow area with a largeradius indicates the elbow. The light gray area is the flatpneumatic actuator when deflated, and the dark gray areais the actuator with inflated. The two pivot points of theactuator play a role of hinges. When the actuator contracts,the initial length of the actuator (Li) shortens and the rotationis made until the length between the two pivot points (Lp)becomes the same as the forearm length (Lf ). If the designfactors, such as the contraction ratio, the arm circumference,etc. are given, any range of motion can be determined basedon simple geometry calculation.

However, the force generated during the contraction of theactuator not only rotates the forearm but also pulls the fore-arm bands in the axial direction. As if this force increases,the forearm band may slip and hinder the operation of theactuator. In order to avoid this unwanted slip, the numberof turns of the actuator has to increase so that the anglebetween the actuator and the forearm band becomes smaller.However, since too many turns could cause discomfort to theuser. a single turn of 360◦ was chosen in our prototype.

The greater the contraction ratio of the actuator, the morethe forearm rotates when the actuators have the same turns.To obtain a large contraction ratio, the length of the singlechamber needs to be wide and long. However, the size of thesingle chamber also determines the entire size of the deviceaffecting the wearability. In reference to previous research,28 mm width and 50 mm length were adopted to achievethe contraction rate of 25% and the wearability, and nine airchambers were chosen [15].

D. Material

Heat-sealable Ripstop fabric (Seattle Fabrics) is a com-posite material of Ripstop fabric and nylon, which has zeroair permeability through the stitches of the fabric. The meritof Ripstop fabric is high burst strength, preventing ripping

Fig. 5. Setup of motorized three-axis heat sealing stage and the expandedview of the heat application unit. The spherical rollerball tip (4.95mm) ismounted to the soldering iron with an adapter made of copper.

caused by frequent inflations. In addition, it can easily beintegrated with other clothing materials by sewing.

The forearm bands were made of double-layered fabric(Polyester 88%, Spandex 12%, 460 g/yd). This fabric hashigh air permeability which enhances the ventilation of thedevice for a long-time use. Moreover, the friction force of thedouble-layered fabric is relatively high because the surfaceof the inner layer is uneven and bumpy. Anti-slip pads werealso attached to the forearm bands for stable adhesion tothe patient’s arm when the device was actuated. Hook-and-loop fasteners and tri-glide buckles were used to fit thedevice to various sizes and geometries of arms. With thesefeatures, easy wearing and removing, stability, comfort andwearability are achieved.

III. FABRICATION

Since reliability, consistency and durability are key factorsin rehabilitation devices, we decided to build an automatedfabrication system for building inflatable actuators with ro-bustness and consistency. There have been different heatsealing methods for fabricating inflatable structures, suchas manual patterning, heat press, automatic patterning [16].Since manual sealing mostly depends on the personal skillsof the operator and does not require a complicated systemfor sealing, it is simple and cost-effective. However, it issometimes time-consuming and does not guarantee the qual-ity of the products. Heat press is a more reliable fabricationmethod for identical products using thermally conductivestencils for sealing. However, it is not be efficient if thesealing patterns of the product changes frequently. Automaticpatterning is the most effective heat sealing method for fast-paced applications compared with the other two methods.In general, a heat applicator is used as an end-effector in amotorized x-y-z stage to draw sealing patterns with simple2D sketches. This method is also free of preparing heavy andbulky stencils. Therefore, we decided to employ automaticpatterning for fabricating inflatable actuators with relativelycomplex patterns.

A. System overview

A soldering iron equipped with a spherical rollerball tipwas mounted on a three axis motorized stage with a work

638

Fig. 6. Definition of the first peak of peeling test. The first peak value ofthe load indicates the quality characteristic value, which is the time whenthe inflatable loses its functionality

space of 950 mm × 900 mm (Fig. 5). The rollerball tip witha diameter of 4.95 mm facilitates smooth movement of theend-effector by minimizing the dragging force while drawingthe pattern. The position in x and y coordinates are controlledby the control unit based on the 2D sketch input. However,the z-coordinate cannot be controlled quickly when there isan unexpected vertical level change of the work space, whichoften happens when the base of the sealing system is notperfectly flat. While the fabrics are not completely sealeddue to the reduced vertical force of the tip on a decliningplane, there is too high normal force exerted to the substrate,resulting in burnt, torn, or squashed fabrics on an inclinedplane. In the worst case, it may cause permanent damage tothe machine too. To address this issue, passive compliance inz-axis was provided for the sealing tip by attaching a linearguide and springs in the mounting block. The linear guideallows only the vertical movement of the end-effector andquickly react to the uneven surface. The pre-tensioned springhelps the end-effector maintain relatively constant pressure tothe fabric substrate in z-axis, which enhances the robustnessand the consistency of the sealing patterns.

B. Characterization

There are a few parameters that determine the quality ofthe sealing patterns: sealing speed, temperature of the heating

TABLE IL9(34) ORTHOGONAL ARRAY FOR THE FIRST TEST SET

Sealing Speed Temperature Z-value S/N RatioTest Sample No.

(mm/min) (◦C) (mm) (dB)#1 150 350 2 22.28#2 150 300 3 13.62#3 150 250 4 13.06#4 100 350 3 23.75#5 100 300 4 18.17#6 100 250 2 10.37#7 50 350 4 24.30#8 50 300 2 19.28#9 50 250 4 10.88

Fig. 7. S/N ratios for (a) the first test set and (b)the second test set. Redcircles indicate the highest S/N ratios based on each parameter

element, and the initial height of the end-effector in z-axis.We used Taguchi methods to find right parameters for robust-ness of the actuator. The initial parameters with three levelswere displayed in the L9(3

4) orthogonal array. We first madenine test samples with a 30 mm long straight line pattern, totest with different parameters as listed in Table I. Then, weconducted peeling tests with the samples with a motorizedtensile test stand (Mark-10, ESM303). The first peak valueof the load is determined to be the quality characteristicvalue because the inflatable actuator loses its functionalityonce it starts leaking (Fig. 6). It means that the sealingrobustness increases as the first peak load becomes larger.Thus, Taguchi’s larger-the-better theorem [17] is applied tocalculate the signal-to-noise (S/N) ratios shown in Eq. (1).

S/N ratio = 10 · log 1n

(i=1∑n

1

y2i

)(1)

where y1 is the quality characteristic value and n is thetest number. The experiment levels with the highest S/Nratios were selected for the second test (Fig. 7-a). Table IIsummarizes the parameters for the second test samples. Thetemperature value was fixed to 350◦C because it mightburn the fabric at a higher temperatures. The orthogonaldesign was changed to L4(2

3) because there were physicallimitations in other parameters. As a result of the experi-ments, the sealing speed, the temperature, and the z-valuewere determined to be 40 mm/min., 350◦C, and 4.5 mm,respectively (Fig. 7-b).

TABLE IIL4(23) ORTHOGONAL ARRAY FOR THE SECOND TEST SET

Z-value Sealing Speed S/N RatioTest Sample No.

(mm) (mm/min) (dB)#1 4 50 20.42#2 4 40 22.92#3 4.5 50 21.80#4 4.5 40 24.51

639

Fig. 8. Force-Length relationship of the device with different pressurelevels

IV. EXPERIMENTS

A. Characterization

Experiments for characterization was conducted to obtainthe relationships between the contraction force and the lengthof the actuator and between the forearm rotation angle andthe torque of the wearable device. In the first experiment,tension between the two ends of the actuator was measuredas the length of the actuator decreased under five differentpressure levels (10, 20, 30, 40, and 50 kPa) with themotorized test stand (Mark-10, ESM303). Similar to otherpneumatic actuators, the contraction force decreased as thelength of the actuator shortened (Fig. 8). In the secondexperiment, a torque sensor (RFT60-HA01, ROBOTUS) andan encoder (SME360CAP-12, SERA) were mounted at thebottom of the aluminum structure and the arm model wasrotated along the axis of the encoder (Fig. 9). There wasno mechanical resistance during rotation, since the arm wasattached only to the axle of the encoder. After fixing thehand to the wrist holder shown in Fig. 9, the actuator wasinflated and the torque was measured by the force-torquesensor at the top of the arm setup. The radius of the upperpart of the arm to which the force was applied was 4.14 cm.The z-axis torque was measured under five different pressurelevels (10, 20, 30, 40, and 50 kPa) as the arm was rotated180◦ with increments of 5◦. The graph shows that the torqueapproached to zero as the arm angle reached 180◦. Since theactuator was designed to rotate up to 180◦, the result showsthe approximation that the two ends of the actuator work asa pivot point was reasonable (Fig. 10).

B. Position Control with Vision Feedback

Position control was tested to evaluate the feasibility ofhelping patients rotate their forearm while they could notreach the desired positions by themselves. With the charac-terization results, pressures for each actuator were calculatedby force equilibrium. Since the data from the characterizationwere discrete, linear interpolation was used. In addition tothis open-loop control, the device was tested with closed-loop control, since the user’s own muscles may disturb thedesired motion of the device.A vision system composed of anRGB and a infrared cameras (Kinect) was set up to collect theangle changes of the forearm. MATLAB was used to process

Fig. 9. Experimental setup for characterization of the device

Fig. 10. Torque vs. rotation angle with varied pressure levels.

the vision data. The vision system was placed on the groundand the arm model was rotated above it (Fig. 11).

The trajectory was planned at a rate of 360◦/min. To reachthe target angle, the two actuators have to reach the forceequilibrium. From the results of the previous experiment, theforce decreases nonlinearly as the rotation angle increases.If the rotation angle of the first actuator is determined, therotation angle of the other actuator could be obtained bysubtracting the angle of the first actuator from 180◦. Then,different air pressures are supplied to the actuators to reachthe force equilibrium.

For closed-loop, circular red and green circle markers wereplaced on the thumb and the little finger of the arm modelrespectively. The sampling rate of the vision data was about6 Hz. The MATLAB recognized the two circle markers aspoints and measured the angle between the horizontal lineand the line made by connecting the two points. Proportionalcontrol was employed for simplicity in this stage.

The result shows that the arm model was able to followthe desired angle. From 40◦ to 140◦, the root-mean-squareerror (RMSE) between the desired angle and the controlledangle was about 7.83◦. A small proportional gain was setin order to avoid oscillation in this experiment. The resultshowed that the controlled angle followed the desired anglein general. However, relatively large errors were shown inthe low and the high angle ranges (0◦-40◦ and 140◦-180◦).When the actuator fully contracted, the actuator generated

640

Fig. 11. Experimental setup for closed-loop control using vision system

Fig. 12. Result of closed-loop angle control of the arm model with thewearable device

a weak torque that could be overcome by the torque of theopposite actuator with little input air pressure. Consideringthis problem, the length of the actuator should be longer.Since the experimental setup is symmetrical, the oppositedirection control can be also carried out in the same way.

V. CONCLUSION AND FUTURE WORK

In this work, we propose a soft wearable robotic devicemade of lightweight fabric inflatable structures, whose designwas inspired by the human musculoskeletal system. Forrobustness and safety, a custom-designed three-axis heatsealing system was built, and its parameters for reliablefabrication were found. In order to verify the feasibilityof the device, experiments on characterization and controlwere conducted. Since the device has the advantages ofcompliance, safety, cost efficiency, and lightweight, it isexpected to be used as an alternative to existing heavy andexpensive rehabilitation devices.

Although silicone pads and high friction material werecurrently chosen for the device, the device often slipped onthe skin when a high air pressure was supplied. To addressthis issue and improve the wearability, clothing techniques,such as wrist guards and elbow supports, will be employedfor easy anchoring to the human body in the future. Since

the proposed device assists only pronation and supinationmotions, other types of upper limb rehabilitation devices willbe integrated into the proposed device as future work.

REFERENCES

[1] S. Hesse, G. Schulte-Tigges, M. Konrad, A. Bardeleben, andC. Werner, “Robot-assisted arm trainer for the passive and activepractice of bilateral forearm and wrist movements in hemipareticsubjects1,” Archives of Physical Medicine and Rehabilitation, vol. 84,no. 6, pp. 915–920, 2003.

[2] A. C. Lo, P. D. Guarino, L. G. Richards, J. K. Haselkorn, G. F.Wittenberg, D. G. Federman, R. J. Ringer, T. H. Wagner, H. I. Krebs,B. T. Volpe et al., “Robot-assisted therapy for long-term upper-limbimpairment after stroke,” New England Journal of Medicine, vol. 362,no. 19, pp. 1772–1783, 2010.

[3] O. Lambercy, L. Dovat, H. Yun, S. K. Wee, C. Kuah, K. Chua,R. Gassert, T. Milner, C. L. Teo, and E. Burdet, “Rehabilitationof grasping and forearm pronation/supination with the haptic knob,”in Rehabilitation Robotics, 2009. ICORR 2009. IEEE InternationalConference on. IEEE, 2009, pp. 22–27.

[4] D. Andreasen, A. Aviles, S. Allen, R. Guthrie, B. Jennings, andS. Sprigle, “Exoskeleton for forearm pronation and supination re-habilitation,” in Engineering in Medicine and Biology Society, 2004.IEMBS’04. 26th Annual International Conference of the IEEE, vol. 1.IEEE, 2004, pp. 2714–2717.

[5] E. Rashedi, A. Mirbagheri, B. Taheri, F. Farahmand, G. Vossoughi, andM. Parnianpour, “Design and development of a hand robotic rehabili-tation device for post stroke patients,” in Engineering in Medicine andBiology Society, 2009. EMBC 2009. Annual International Conferenceof the IEEE. IEEE, 2009, pp. 5026–5029.

[6] S. Ito, H. Kawasaki, Y. Ishigure, M. Natsume, T. Mouri, and Y. Nishi-moto, “A design of fine motion assist equipment for disabled hand inrobotic rehabilitation system,” Journal of the Franklin Institute, vol.348, no. 1, pp. 79–89, 2011.

[7] H. S. Lo and S. Q. Xie, “Exoskeleton robots for upper-limb rehabili-tation: State of the art and future prospects,” Medical Engineering &Physics, vol. 34, no. 3, pp. 261–268, 2012.

[8] S. Kyeong, G. M. Gu, and J. Kim, “Development of a portable forearmmirror image rehabilitation device,” in Ubiquitous Robots and AmbientIntelligence (URAI), 2015 12th International Conference on. IEEE,2015, pp. 417–418.

[9] S.-S. Yun, B. B. Kang, and K.-J. Cho, “Exo-glove pm: an easilycustomizable modularized pneumatic assistive glove,” IEEE Roboticsand Automation Letters, vol. 2, no. 3, pp. 1725–1732, 2017.

[10] Y.-L. Park, B.-r. Chen, N. O. Perez-Arancibia, D. Young, L. Stirling,R. J. Wood, E. C. Goldfield, and R. Nagpal, “Design and control of abio-inspired soft wearable robotic device for ankle–foot rehabilitation,”Bioinspiration & Biomimetics, vol. 9, no. 1, p. 016007, 2014.

[11] V. Oguntosin, W. S. Harwin, S. Kawamura, S. J. Nasuto, andY. Hayashi, “Development of a wearable assistive soft robotic devicefor elbow rehabilitation,” in Rehabilitation Robotics (ICORR), 2015IEEE International Conference on. IEEE, 2015, pp. 747–752.

[12] R. Niiyama, X. Sun, C. Sung, B. An, D. Rus, and S. Kim, “Pouchmotors: Printable soft actuators integrated with computational design,”Soft Robotics, vol. 2, no. 2, pp. 59–70, 2015.

[13] J. Wirekoh and Y.-L. Park, “Design of flat pneumatic artificial mus-cles,” Smart Materials and Structures, vol. 26, no. 3, p. 035009, 2017.

[14] H. Shaaban, C. Pereira, R. Williams, and V. Lees, “The effect of elbowposition on the range of supination and pronation of the forearm,”Journal of Hand Surgery (European Volume), vol. 33, no. 1, pp. 3–8,2008.

[15] R. Niiyama, C. Rognon, and Y. Kuniyoshi, “Printable pneumaticartificial muscles for anatomy-based humanoid robots,” in HumanoidRobots (Humanoids), 2015 IEEE-RAS 15th International Conferenceon. IEEE, 2015, pp. 401–406.

[16] J. Ou, M. Skouras, N. Vlavianos, F. Heibeck, C.-Y. Cheng, J. Pe-ters, and H. Ishii, “aeromorph-heat-sealing inflatable shape-changematerials for interaction design,” in Proceedings of the 29th AnnualSymposium on User Interface Software and Technology. ACM, 2016,pp. 121–132.

[17] G. S. Peace, Taguchi methods: a hands-on approach. Reading, MA:Addison-Wesley.

641