Development of a wearable assistive softrobotic device for elbow rehabilitationConference or Workshop ItemAccepted VersionOguntosin, V., Harwin, W. S., Kawamura, S., Nasuto, S. J. andHayashi, Y. (2015) Development of a wearable assistive softrobotic device for elbow rehabilitation. In: The 2015 ICORR11th International Conference on Rehabilitation Robotics, 1114 Aug 2015, Singapore, pp. 747-752. Available at is advisable to refer to the publisher’s version if you intend to cite from thework. See Guidance on citing .Published version at: outputs in CentAUR are protected by Intellectual Property Rights law,including copyright law. Copyright and IPR is retained by the creators or othercopyright holders. Terms and conditions for use of this material are defined inthe End User Agreement . Archive at the University of ReadingReading’s research outputs online

Development of a Wearable Assistive Soft Robotic Device for ElbowRehabilitationVictoria Oguntosin, William S. Harwin, Sadao Kawamura, Slawomir J. Nasuto and Yoshikatsu HayashiAbstract— The loss of motor function at the elbow joint canresult as a consequence of stroke. Stroke is a clinical illnessresulting in long lasting neurological deficits often affectingsomatosensory and motor cortices. More than half of those thatrecover from a stroke survive with disability in their upper armand need rehabilitation therapy to help in regaining functionsof daily living. In this paper, we demonstrated a prototypeof a low-cost, ultra-light and wearable soft robotic assistivedevice that could aid administration of elbow motion therapiesto stroke patients. In order to assist the rotation of the elbowjoint, the soft modules which consist of soft wedge-like cellularunits was inflated by air to produce torque at the elbow joint.Highly compliant rotation can be naturally realised by theelastic property of soft silicone and pneumatic control of air.Based on the direct visual-actuation control, a higher controlloop utilised visual processing to apply positional control, thelower control loop was implemented by an electronic circuit toachieve the desired pressure of the soft modules by Pulse WidthModulation. To examine the functionality of the proposed softmodular system, we used an anatomical model of the upperlimb and performed the experiments with healthy participants.I. INTRODUCTIONStroke results from damage to the vascular supply of thebrain resulting in neuronal death in the brain [1]. Therefore,stroke patients have difficulty performing specific motionssuch as elbow flexion and extension due to damage of themotor and somatosensory cortices. Early muscle activationis critical to good recovery [2], hence the need for exercises[1], [3]. Stroke neuro-rehabilitation is therefore importantto restore muscle functions to the damaged body parts andhelp stroke surviving patients regain the ability to performactivities of daily living. This rehabilitation is requiredespecially during the early stage of stroke diagnosis whenthere is a greater opportunity to influence neural plasticityand brain recovery [4]. Stroke rehabilitation requires long,manipulative and intensive direct physical therapy sessions toimprove strength, accuracy, and range of motion [5]. Recentresearch work on robot therapies have been shown to besafe and beneficial over human delivered therapies due totheir flexibility and intensity [6]. An inexpensive, compliant,wearable and lightweight robotic device allowing patientsto engage in exercises on their own, either at home or inVictoria Oguntosin, William S. Harwin, Slawomir J. Nasuto andYoshikatsu Hayashi are with Systems Neuroscience Group, School ofSystems Engineering, University of Reading, U.K. Sadao Kawamurais with Department of Robotics, Ritsumeikan University, Japan. Email:;[email protected]; [email protected] hospital, would make physical therapy more available topatients [3], [7].There has been a growing interest in using soft actuatorsfor rehabilitation training, for example, the McKibben pneumatic artificial muscles are the most highly developed andstudied class of soft actuators [8], [9]. State of the art inrehabilitation soft robotics includes lower limb gait rehabilitation [10] to facilitate hip movement for a rodent with spinalcord injury, a rehabilitation glove to achieve flexing motionfor the thumb [11] and human fingers [7] by generatingsufficient amount of speed, range of motion and forces forthese motions. Fluidic driven upper limb orthosis have beenused for rehabilitation [12]. Soft robots are predominantlymade of continuously deformable high-strain rubber [13].This feature makes them inherently soft and therefore safefor direct attachment to the skin. Pneumatic actuation is oftenrequired to generate assistive torques to joints of the body.Soft robots have been shown to deliver motions rangingfrom single degree-of-freedom movement such as bendingand angular displacement to multiple modes of gait [14],[15], [16], [17]. The ease of fabrication and low cost ofmanufacture and actuation [18] enables actuator geometrydesign and production in a short time.In contrast to conventional assistive devices [19], [20]which are made of rigid heavy metal, and as a result, requirethe constrain of the body trunk, we aim at wearable andlightweight assistive modules which can provide appropriatedegrees of assistive force to the elbow. The lightweightmodules made of soft inflatable structures allow us to provideassistive force while ensuring safe human-robotic interaction[8] required for neuro-rehabilitation. In contrast to the stateof-the art of soft robots currently used to provide assistivemotion, our design utilises a low air pressure supply of50KPa, with highly compliant silicone rubber having shoreA hardness. The cost of the complete soft robotic deviceincluding the actuation system is extremely low and veryaffordable.We describe the design of an entirely soft and inexpensiverobot, composed of 100% silicone rubber which offers upperlimb assistance for the elbow. The actuator is designed to bewearable around the arm and forearm using silicone bandsas attachment thereby demonstrating its safety, ease of fittingand removal as well as its practicability as a physiotherapy apparatus. The main contribution of this paper is theconstruction and testing of an entirely soft, light-weight,one degree of freedom prototype elbow joint rehabilitation

training system. This system offers upper limb assistance byproviding for elbow flexion and extension. Pneumatic controlof air in the soft modules is achieved by using a pressuresensor as feedback while the angle of motion was obtainedusing visual information from a webcam.III. ACTUATOR DESIGN AND FABRICATIONA. Actuator Geometrical ParametersII. DESIGN CONSIDERATIONSDuring the intensive therapies of the upper limb deliveredto stroke patients by trained physiotherapist, muscles arestretched and the elbow joint is flexed and extended throughits range of motion in order to recover the smooth and elasticfunctions of muscles. Hard robotic structures are mostlyused to produce these motions. They are based on rigidlinks connected by joints - this makes them heavy withexpensive and complicated control. Furthermore, they tendto be supported by a solid base on the ground. However, thiskind of constraint largely restricts the horizon of assistivedevices. Moreover, hard assistive devices require appropriategravity compensation in proportion to the weight of the upperlimb to offset gravity effects. All of these considerations leadto a set of design specifications required to be fulfilled by asoft wearable robotic assistive device in order to successfullyassist in elbow motion.1) Visco-elastic property for passive motion: To realiseideal trajectories, positional control is applied, and theamount of force to restore the desired trajectory should becontrolled compliantly by visco-elastic property of the softactuator.2) Structural transparency for active motion: The structural transparency is such that the mass of the actuator isso small that a patient performing spontaneous motion doesnot feel resistance in interactions with the robotic arm. Anactuator mass less than 0.5kg is desirable [7].3) Wearable assistive robot: It should not require the basecoordinate system of robot to be grounded so that it does notconstrain the motion of the main body.A soft assistive device should be able to deliver appropriateassistive torque for the particular motion according to thetorque normally required for a typical healthy individual. Theproduction of an appropriate assistive torque for rotationalmotion can be difficult due to inherent softness and its abilityto generate a fraction of the normal assistive torque producedby conventional robots.Our design is such that soft wedge-like inflatable modulesare actuated to produce torque. The measured maximumtorque for this actuator is 2.7Nm, the typical human isometricstrength for elbow flexion/extension is 1.2Nm [21] whichproves the viability of the soft modules to assist in thesemotions. When worn on the upper limb and actuated viainflation, a change in wedge angle would result which acts togenerate a rotational motion, thereby assisting the movementof the elbow for extension and flexion while the controlof air pressure in individual modules will contribute to thecompliant assistance. Other specifications such as intrinsicsafety; range of motion; comfort of wearing as well asextremely low manufacturing cost was implemented.Fig. 1: Schematic of a single soft module.In this study, a number of geometrical parameters in thedesign of the soft assistive device can be identified as shownin Figure 1. These parameters are defined by the unactuatedangle of the actuator, base length, height of air channel, widthof the actuator and wall thickness.TABLE I: Basic Characteristics of the Rehabilitation SoftModulesCharacteristicsWeightDegrees of FreedomAssistive TorqueRange of MotionOperating PressureSpecified PurposeSafety factorsValue0.35Kg1 (Elbow Flexion and Extension)2.7Nm90 -130 0 - 50KPaElbow RehabilitationComposed of 100% siliconeEase of wearingThis rehabilitation actuator is composed of four wedgeshaped segment having an individual angle of 20 . Theopposite layers of the 20 angle are made of highly inextensible silicone and the other layers are casted with highstrain silicone. These layers direct the structure to increasethe angle between the top and the bottom layers as theactuator is pressurised; and decrease its angle when depressurised. Angular displacement is therefore achieved bypneumatically pressurizing/depressurizing the embedded airchannels within the soft modules. When pressurised, thechannels undergo high strain and effectively increase inangle. Because of the strain-limiting layers, this expansiongenerates angular displacement about the stiffer layer. Thisdesign is modular by enabling scalability in the number ofwedge-shaped segments. Table I shows a summary of thedesign characteristics of the soft modules.B. Fabrication Process of Soft ActuatorThe wedge-shaped modules were produced by a two-partcasting process. The geometry of the mould for the softactuator was designed using SolidWorks, a 3D CAD designsoftware. The mould is then printed out with Acrylonitrile

(a)(b)(c)Fig. 2: Soft modules worn by a healthy subject. The red, blue and green colour markers were used by the visual processingsystem to measure the current angle of the elbow. a) the soft modules in an unactuated state b) the soft modules aredepressuzied at a vacuum pressure of -10KPa with the elbow in flexion c) the soft modules pressurised at a pressure of7KPa with the elbow in an extended state.Butadiene Styrene (ABS) plastic using HP DeskJet R 3Dprinter. Ecoflex 0030 silicone with a viscousity of 3000cPswas used as the highly-extensible layer while Addition Cure33 silicone with a viscousity of 7000cPs was used as the lowstrain layer. Both grades of silicone liquid comes in two parts- Parts A and B. Both parts were mixed in equal quantity (byweight) and poured into the 3D printed moulds, and allowedto cure at room temperature for 4 hours.The main body of the actuator, composed of the softercured silicone is then immersed into the base mould containing the high-stiffness liquid silicone to create a perfectseal with the bottom layer after curing. These two layers (softmain body and hard base) produces a perfect seal becausecured silicone bonds sufficiently well with liquid siliconeafter curing. After the two layers are bonded together, apuncture size of about 2mm is created to make a tube openinginto the air channel; silicone tubing is then inserted intothe air channel. Lastly, the contact point of the structureand tubing are sealed with a final coating of silicone tominimise air leakage. Repeating these fabrication procedures,four of these same actuators were produced. They were gluedtogether with fine coating of liquid silicone to produce thefinal design.C. Design of AttachmentAttachments of the soft modules to the human upper limbwere designed to facilitate large area of contact, stable fit,comfort of wearing and removal. Attachment was requiredto provide a smooth distribution of contact pressure over theentire length of the arm and forearm thereby minimizingdiscomfort and provide a soft but secure attachment. Thesewere taken into consideration since the soft modules aredesigned for use by individuals with little or no muscularpower. Figure 2 shows the silicone modules attached to ahuman participant.Generally, both the actuator geometry and attachmentshould be designed to suit the geometry of the patient. Inour case, a generic design was made for the purpose ofexperiments.IV. CONTROL SCHEMEThe experimental system of pressure and angle measurement has four major parts (Figure 3): A webcam, computerfor visual processing, pressure regulation system and thewedge-shaped rubber actuator. The webcam provides visualinformation, which is processed by an image processingsoftware. The pressure regulation system consists of a pump(air and vacuum), by which pneumatic pressurization andde-pressurization can be obtained; inlet and exhaust solenoidvalves; pressure sensor; and microcontroller.Fig. 3: Schematic picture of experimental system. The softmodules are attached to the elbow joint with red, blue andgreen colour markers positioned as shown, the Visual C program returns the angle of the elbow joint, the pressuresensor provides fedback by measuring the air pressure insidethe modules.A. Visual ProcessingVisual processing was implemented in Visual C usingOpenCV libraries. The processing was carried out in HSV(Hue-Saturation-Value) space. The video frames were preprocessed to filter and reduce noise. Red, Blue and Greencolour markers were used to track the position of the upperarm and forearm. The centroid of each of the colour markersdefined by their x and y coordinates were obtained as the

(a)(b)(c)Fig. 4: Pressure of air and measured angle of extension/flexion of elbow while the soft modules were tested on a typicalhealthy test subject. The angle measurements were obtained via visual processing while the pressure data was obtained fromreadings from an analog gauge pressure sensor. a) Measured range of angle of elbow extension as a function of time. b)Measured range of angle of elbow flexion as a function of time. c) Pressure of air in the soft modules as a function of timewhile the modules performed extension motion.position of the marker. Figure 5 shows the camera view ofthe image. The angle of the elbow is calculated as:q0 cos 1 (l12 l22 l32)2l1 l2(1)Where lengths l1 , l2 and l3 are calculated as the distance (inunit of pixels) between the centroid of the red and blue; blueand green; red and green colour markers respectively.established through the serial interface of the computer asshown in Figure 3.C. Pressure Regulation CircuitFigure 6 shows the arrangement of pumps and valves forthe soft actuation of the modules. As shown, it consists of anair pump to act as the pressure source; two solenoid valves(one acts as an inlet valve while the other as an exhaustvalve) and an amplified analog pressure sensor to measurethe air pressure in the soft modules. The solenoid valves actto control the flow of air into and out of the soft modules. Theelectronic circuit consists of a PIC16F876A microcontrollerrunning at a clock speed of 4MHz and transistors to enablethe high current valves and pump to be controlled by themicrocontroller as shown in Figure 7. The solenoid valvesare controlled by 2 output compare modules configured forPWM-mode. As no operating system kernel is used, mainlyinterrupt sub-routines and parallel running subunits are used.This works very well for analog input of pressure signals.Fig. 5: The red, blue and green markers are positioned asindicated for measurement of the current elbow angle, q0 .The webcam projects a 2D image of the scene as shown.Each x, y position represents the pixel co-ordinate [x, y] ofthe center position of each colour marker.B. Interfacing the microcontroller with the visual processingprogramA serial communication link was established to obtain thecurrent angle measurement q0 from the Visual processingprogram to the microcontroller which in turn would generatean appropriate Pulse Width Modulated (PWM) signal thatwould actuate the pump and valves to control the air pressureinto the soft modules. This communication pathway wasFig. 6: Circuit arrangement of pressure controller. Figureshows the arrangement of the air-inlet and exhaust valves,pump, pressure sensor and soft modules.Compressed air at a suppy pressure of 50KPa from the

air pump passes through the air supply solenoid valve andchanges to output pressure when the air supply solenoidvalve turns ON. Therefore, air from the supply pump passesthrough the air supply solenoid valve and changes to outputpressure. A PWM output is then produced on the outputpin of the microcontroller to switch ON/OFF the exhaustvalve in order to produce an output pressure equal to thedesired pressure. The output pressure is fed back to themicrocontroller via the pressure sensor. This is to check ifthe desired pressure, Pd , has become equal to the outputpressure P0 . Pressure corrections then occur to produce anoutput pressure that is equal to the set pressure.setpoint to send the appropriate PWM signal to the solenoidvalves to regulate the pressure in the soft actuator. Figure8 shows the block diagram for the joint visual and pressurecontrol. With this control method, open loop joint controlis possible by varying the amount of desired pressure, Pd ,that is added to the output of the control loop. The pressurefeedback signal is provided by the means of a pressure sensorto measure pressure in the soft modules. The output of theseinner pressure control corresponds to the duty cycle of thePWM signal that drives the solenoid valves. To measurepressure, amplified pressure sensors (0 - 5psi pressure range)with analog interface having a 1ms response time was used.A 10-bit Analog-to-Digital Converter (ADC) resolution wasused to read the pressure. Using the microcontroller PWMfeatures, pressure is controlled by varying the length of theduty cycle of the air pump.V. METHODSFig. 7: Control circuit for the soft modules. Figure shows thesoft modules, air-inlet and exhaust solenoid valves, air pump,transistors, pressure sensor and the PIC microcontroller.D. Joint ControlA light weight anatomical model of an elbow joint madeof swimming float and 3D-printed hinge joint was first usedto test the use of these soft modules. The model was usedas a first step in taking measurements. Thereafter healthysubjects were used to verify the use of the soft modules.A typical subject sits on a chair, puts on the soft modules,places the elbow on a table while exercises are carried outwith pneumatic actuation. The assistive soft modules wastested on three healthy participants between the ages of25 5 yrs and body weight within the range of 65 10kg.The participants wore the soft modules for a duration of 15minutes while actuation was carried out and measurementdata were recorded. Figure 2 shows the modules rotating theelbow of a test subject.VI. EXPERIMENTAL RESULTSFig. 8: Block Diagram of Joint Control. The lower pressurecontrol loop uses information from the pressure sensor toget the output pressure, P0 while the higher visual controlreceives information of the current angle of the elbow, q0from the webcam.A controller for visual feedback provided by the webcamand pressure control are required for joint control. Thisvisual-pressure control loop periodically receives discrepancies between the measured elbow angle q0 and desired angleqd and uses a two-staged cascaded control approach to senda PWM signal to the pump and solenoid valves to resolvethe difference. The time required for the computer visionprogram to process each frame was measured as 200ms.The difference between the measured and the referenceangle results in an error signal, which when passed to theproportional controller outputs a signal which is then used asThe graphs for angle and air pressure measurements fora typical test participant is shown in Figure 4. The experiments were conducted by using the soft modules to actuatethe elbow while pressure and angle measurements wererecorded. The purpose of these experiments was to showthat through a combination of the control (visual processingand microcontroller program) and design (that is the inherentsoftness, lightness and compliance of the soft modules), wecan successfully accomplish the primary task of actuating anelbow joint for flexion and extension.Figure 4c shows the pressure of air in the soft modulesduring actuation. Measurements of air pressure in the softactuator was taken with an analog pressure sensor. When anindividual triangle segment is inflated, there is a change inangle and also a bulge on the soft layer of the triangle. This isbecause the inflation pressure P0 gives rise to extension boththe circumferential and axial directions. This bulge reducedthe efficiency at which the triangle modules increased inangle. If the bulge can be minimised, the angle change ineach of the triangle segment would increase.Using 6V DC supply to power the control circuit at asupply pressure of 50KPa, and for a desired elbow flexion

and extension motion between 70 - 110 , the measuredmaximum frequency of the soft modules is 0.03Hz. For thissystem, it can be difficult to achieve high accuracy of visionand pressure measurements. This is expected as a result ofdelay by the high-level visual control and also due to someparameters of the solenoid valve such as leakage and exhaustspeed. Also, air leakage at the connecting point between thesoft robot and the tubing; loose tubing connection betweenthe pump, valves and pressure sensor will limit the accuracyof this system. In order to achieve a good computer visionmeasurement system, a recalibration of the HSV values ofthe three colour markers was done every time measurementswere taken. This was done so as to accurately track thecorrect pixel coordinate of the colour markers each timethe experiment was conducted, thereby minimising the effectof change in lighting conditions of the room and ambientlight. The second area of performance evaluation is the jointoutput torque the soft modules can produce. The measuredmaximum torque achieved for this rehabilitation device is2.7Nm.VII. CONCLUSIONA wearable assistive rehabilitation device for stroke patients or individuals with little or no power/muscle controlin the elbow was designed. To achieve this, an entirely softdesign constructed uniquely from only silicone rubber, withinherent low weight and safety was implemented. The abilityof this entirely soft assistive arm to rotate an elbow joint wasdemonstrated, which leads to many potential applicationssuch as providing muscle power, restoring motor functionsand as a neuro-rehabilitation device. A microcontroller operated pressure regulator system to perform pneumatic actuation of the soft modules was also presented. A visualprocessing system to record the angle measurements of theelbow was presented. The system was designed to meet thespecifications of low weight, ease of wearing and comparatively low mechanical complexity, which are essential for anydevice adapted for use on the human body. The performanceof the soft modules suggests its application as a physiotherapy training device for elbow flexion and extension. Thecomplete soft robotic device including the cost of the softmodules and actuation components is essentially low andtherefore can be readily available to every member of thepopulace including developing countries.ACKNOWLEDGMENTThe authors would like to thank The Great BritainSasakawa Foundation (Project Number: 4391) for the international exchange, Dr. Guy Haworth for his support, Mr. BenHaworth for his master thesis, Mr. Steve Gould and Mr. NickDove for their help in producing the control board.R EFERENCES[1] W. S. Harwin, A. Murgia and E. K. Stokes, ”Assessing the effectiveness of robot facilitated neurorehabilitation for relearning motorskills following a stroke”, Medical and Biological Engineering andComputing, Springer 49 No.10, pp. 1093-1102, 2011.[2] Y. Hayashi, K. Nagai, K. Ito, J. S. Nasuto, C.V. R. Loureiro and S.W. Harwin, ”Feasible Study of EEG driven Assist Robot System forStroke Rehabilitation”, IEEE International Conference on BiomedicalRobotics and Biomechatronics, pp. 1733-1739, 2002.[3] S. Coote, B. Murphy, W. Harwin and E. Stokes, ”The effect ofthe GENTLE/s robot-mediated therapy system on arm function afterstroke”, Clinical Rehabilitation, No.5, pp. 395-405, 2008.[4] R. C. V. Loureiro and W. S. Harwin, ”Reach & grasp therapy:Design and control of a 9-DOF robotic neuro-rehabilitation system”,International Conference on Rehabilitation Robotics, pp. 757-763,2007.[5] C. D. Takahashi, L. Der-Yeghiaian, V. H. Le, and S. C. Cramer,”A Robotic Device for Hand Motor Therapy After Stroke”, IEEEInternational Conference on Rehabilitation Robotics, 2005.[6] R. C. V. Loureiro, W. S. Harwin, R. Lamperd and C. Collin, ”Evaluation of Reach and Grasp Robot-Assisted Therapy Suggests SimilarFunctional Recovery Patterns on Proximal and Distal Arm Segmentsin Sub-Acute Hemiplegia”, IEEE Transactions on Neural Systems andRehabilitation Engineering, pp. 593-602, 2014.[7] P. Polygerinos, S. Lyne, Z. Wang, F. L. Nicolini, B. Mosadegh, G.M. Whitesides and C. J. Walsh, ”Towards a Soft Pneumatic Glove forHand Rehabilitation”, IEEE International Conference on IntelligentRobots and Systems, 2013.[8] N. G. Tsagarakis and D. G. Caldwell, ”Improved modelling and assessment of pneumatic muscle actuators”, IEEE International Conferenceon Robotic and Automation, pp. 3641-3646, 2000.[9] B. Tondu, P. Lopez, ”Modeling and control of McKibben artificialmuscle robot actuators”, IEEE Control Systems, Vol. 20, Issue 2, 2002.[10] Y. S. Song, Y. Sun, R. Brand, J. Zitzewitz, S. Micera, G. Courtine andJ. Paik, ”Soft Robot for Gait Rehabilitation of Spinalized Rodents”,IEEE International Conference on Intelligent Robots and Systems,2013.[11] P. Maeder-York, T. Clites, E. Boggs, R. Neff, P. Polygerinos, D. Holland, L. Stirling, K. C. Galloway, C. Wee, C. J. Walsh, ”BiologicallyInspired Soft Robot for Thumb Rehabilitation”, Proceedings of Designof Medical Devices Conference, 2014.[12] S. Schulz, B. Schmitz, R. Wiegand, C. Pylatiuk and M. Reischl,”The Hybrid Fluidic Driven Upper Limb Orthosis-OrthoJacket”, Proceedings of the MyoElectric Controls/Powered Prosthetics SymposiumFredericton, 2011.[13] A. D. Marchese, R. K. Katzschmann and D. Rus, ”Whole ArmPlanning for a Soft and Highly Compliant 2D Robotic Manipulator”,IEEE International Conference on Intelligent Robots and Systems, pp554 - 560, 2014.[14] R. F. Shepherd, F. Ilievskia, W. Choia, S. A. Morina, A. A. Stokes, A.D. Mazzeoa, X. Chena, M. Wanga and G. M. Whitesides, ”Multi-gaitsoft robot”, PNAS, 2011.[15] F. Ilievski, A. D. Mazzeo, R. F. Shepherd, X. Chen, and G.M.Whitesides, ”Soft Robotics for Chemists”, Angewandte Chemie, Vol.123, pp. 1930 - 1935, 2011.[16] R. V. Martinez, J. L. Branch, C. R. Fish, L. Jin, R. F. Shepherd,M. D. Nunes, Z. Suo and G.M. Whitesides, ”Robotic Tentacles withThree-Dimensional Mobility Based on Flexible Elastomers”, AdvancedMaterials, Vol. 25, pp 205 212, 2013.[17] Y. Sun, Y. S. Song and J. Paik ”Characterization of Silicone RubberBased Soft Pneumatic Actuators”, IEEE International Conference onIntelligent Robots and Systems, 2013.[18] V. Oguntosin, S. J. Nasuto and Y. Hayashi, ”A Compact Low-CostElectronic Hardware Design for Actuating Soft Robots”, UKSIMAMSS International Conference on Modelling and Simulation, pp. 242247, 2015.[19] N. G. Kutner, R. Zhang, A. J. Butler, S. L. Wolf and J. L. Alberts,”Quality-of-life change associated with robotic-assisted therapy toimprove hand motor function in patients with subacute stroke: arandomized clinical trial”, Physical therapy, Vol. 90, pp. 493- 504,2010.[20] H. I. Krebs, B.T. Volpe, M.L. Aisen and N. Hogan, ”Increasingproductivity and quality of care: Robot-aided neuro-rehabilitation”,Journal of Rehabilitation Research and Development, 2000.[21] A. Gil-Agudo, A. Ama-Espinosa, A. Reyes-Guzmn, A. Bernal-Sahunand E. Rocon, ”Applications of Upper Limb Biomechanical Modelsin Spinal Cord Injury Patients”, Biomechanics in Applications, 2011.

robotic device for elbow rehabilitation Conference or Workshop Item Accepted Version Oguntosin, V., Harwin, W. S., Kawamura, S., Nasuto, S. J. and Hayashi, Y. (2015) Development of a wearable assistive soft robotic device for elbow rehabilitation. In: The 2015 ICORR 11th International Conference on Rehabilitation Robotics, 11-